The integration of artificial skin into robotic systems has long been foreseen. For the last decades the introduction was always just a couple of years away. One of the possible reasons for the slow progress is the focus of the research efforts on high sensitivity and spatial resolution. The resulting tactile sensor prototypes are presented as laboratory prototypes with no or little chance to be successfully integrated into a robotic system. In order to enable the operation of an artificial skin on a robotic system in real-world applications the conflict of goals between high sensitivity and an overload-proof, collision tolerant, design needs to be solved. Within this paper we present a stretchable tactile surface sensor as a major functional component of an artificial skin setup that is able to unite the required sensitivity with the ability to withstand collisions. The results of first experiments regarding sensitivity and the ability of the DLR Artificial Skin setup to withstand high impact forces are presented.

This paper introduces the new robot programming language LightRocks(Light Weight Robot Coding for Skills), a domain specific language (DSL) for robot programming. The language offers three different level of abstraction for robot programming. On lowest level skills are coded by domain experts. On a more abstract level these skills are supposed to be combined by shop floor workers or technicians to define tasks. The language is designed to allow as much flexibility as necessary on the lowest level of abstraction and is kept as simple as possible with the more abstract layers. A Statechart like model is used to describe the different levels of detail. For this we apply the UML/P and the language workbench MontiCore. To this end we are able to generate code while hiding controller specific implementation details. In addition the development in LightRocks is supported by a generic graphical editor implemented as an Eclipse plugin.

Feedback of interaction forces/torques in robot supported minimally invasive surgery (MIRS) meanwhile is a widely accepted advantage. To avoid parasitic effects on the according force/torque sensor, e. g. due to friction in the trocar, a sensor integration as distal as possible is advisable. Since common sensor principles are based on wire or fibre connection and since an axisymmetrical instrument's distal end has to be actuated in at least one additional degree of freedom (DoF) for full dexterity inside the patient, cable (or fibre) break is a serious problem. A constraint guided spine-kinematics without rotatory DoF at its distal end is proposed to reduce the danger of cable (or fibre) break due to a curved flection instead of a sharply bending joint. The properties of a constraint spine-kinematics with one DoF adapted to the special requests of an in-house developed MIRS-system is presented, verification data acquired in a specially developed test bed are shown. The results seem suitable for MIRS applications and for a progress towards a 2 DoF solution with the presented constraint principle.

The grasping and stabilization of a spinning, noncooperative target satellite by means of a free-flying robot is addressed. A method for computing feasible robot trajectories for grasping a target with known geometry in a useful time is presented, based on nonlinear optimization and a look-up table. An off-line computation provides a data base for a mapping between a four-dimensional input space, to characterize the target motion, and an N-dimensional output space, representing the family of time-parameterized optimal robot trajectories. Simulation results show the effectiveness of the data base for computing grasping maneuvers in a useful time, for a sample range of spinning motions. The debris object consists of a satellite with solar appendages in Low Earth Orbit, which presents collision avoidance and timing challenges for executing the task.

The technological progress in the field of robotics results in more and more complex manipulators. However, having an increasing number of degrees of freedom raises the question of how to use them effectively. In turn, establishing manipulators in human environments, e.g., as service robots, calls for the fulfillment of various constraints and tasks at the same time. In the context of torque controlled robotic systems, we provide an approach to simultaneously deal with a multitude of tasks and constraints which are arranged in a hierarchy, utilizing the large number of actuated joints of the manipulator. To this end, we propose a continuous null space projection technique to consider unilateral constraints, singular Jacobian matrices and dynamic variations of the priority order within the hierarchical structure. We show that activating and deactivating tasks as well as crossing singularities does not lead to a discontinuous control law. Simulations and experiments on the humanoid Justin of the German Aerospace Center (DLR) validate our approach. The presented concept is supposed to contribute to whole-body control frameworks.

This paper presents the identification of contact formations using force torque information. As force torque measurements do not map uniquely to their corresponding contact formations, three steps are performed: Initially, the wrench space for each contact formation is computed automatically. Then, a contact formation graph is augmented with a similarity index that reflects the similarity of contact formations with respect to their spanned wrench spaces. A particle filter is used to represent the likeliness of a contact formation given a force torque measurement. Finally, this probability distribution is resolved taking the similarity index, the transitions of the contact formation graph and the history of identified contact formations into account. This allows the recognition of the order of demonstrated contact formations by a measured set of forces and torques. The approach is verified by experiments.

A different way of force control is presented, that is especially advantageous for position controlled robots. Instead of usual force control laws we rely on the well tuned position control loop and just use the force sensor to measure the target pose or to predict the desired trajectory. In combination with a compliant sensor we introduce an inherently stable framework of force control which almost inhibits all control errors. After an unexpected impact the force error is reduced independently from the sensor's bandwidth or delays in signal processing. Thus the (inevitable) impact force is more significant than the measured force control errors. The special case of a sensor that is mounted far away from a vertex-face contact is discussed, too.

In this paper we adressed the problem of finding exciting trajectories for the identification of manipulator link inertia parameters. This can be formulated as a constraint nonlinear optimization problem. The new approach in the presented method is the parameterization of the trajectories with optimized B-splines. Experiments are carried out on a 7 joint Light-Weight robot with torque sensoring in each joint. Thus, unmodeled joint friction and noisy motor current measurements must not be taken into account. The estimated dynamic model is verified on a different validation trajectory. The results show a clear improvement of the estimated dynamic model compared to a CAD-valued model.

This paper presents an approach for computing power grasps for hands with kinematic structure similar to the human hand, which allows the implementation of strategies inspired in human grasping actions. The proposed method first samples the object surface to look for the best spots for creating an opposing grasp with two or three fingers, and then aligns the other fingers to match the local curvature of the object surface. Different grasp strategies are considered, depending on the relative size of the object with respect to the hand, and on the location of potential obstacles in the environment. Several application examples are provided with two different hand models.

We present a next-best-scan (NBS) planning approach for autonomous 3D modeling. The system successively completes a 3D model from complex shaped objects by iteratively selecting a NBS based on previously acquired data. For this purpose, new range data is accumulated in-the-loop into a 3D surface (streaming reconstruction) and new continuous scan paths along the estimated surface trend are generated. Further, the space around the object is explored using a probabilistic exploration approach that considers sensor uncertainty. This allows for collision free path planning in order to completely scan unknown objects. For each scan path, the expected information gain is determined and the best path is selected as NBS. The presented NBS approach is tested with a laser striper system, attached to an industrial robot. The results are compared to state-of-the-art next-best-view methods. Our results show promising performance with respect to completeness, quality and scan time.

This paper presents an optical-inertial tracking algorithm with explicit and assured optical marker identification. Active optical markers are sequentially or simultaneously triggered to achieve a maximum in quantity and quality of measurements available for tracking. Markers that appear in a camera image are identified by individual activation and are locally tracked in the 2D-images after initial identification. The 2D-position measurements of the cameras are combined with low latency measurements of acceleration and angular velocity from an inertial measurement unit. An Extended Kalman Filter is used for an ultra-tightly coupled data fusion, that takes advantage of all marker measurements with verified identity. The accurate, low latency tracking is robust with respect to temporary marker occlusions, as needed in applications where a robot is directly controlled with a tracked device. The tracking algorithms are implemented in real-time and verified with a test bed in a medical robotics context.

This work presents an evaluation study of two different collision feedback modalities for virtual assembly verification: visual and force feedback. Forty-three subjects performed several assembly tasks (peg-in-hole, narrow passage) designed with two levels of difficulty. The used haptic rendering algorithm is based on voxel and point data-structures. Both objective - time and collision performance - and subjective measures have been recorded and analyzed. The comparison of the feedback modalities revealed a clear and highly significant superiority of force feedback in virtual assembly scenarios. The objective data shows that whereas the assembly time is similar in most cases for both conditions, force collision feedback yields significantly smaller collision forces, which indicate higher assembly precision. The subjective ratings of the participants define the force feedback condition as the most appropriate for determining clearances and correcting collision configurations, being the best suited modality to predict mountability.

Based on HIT/DLR (Harbin Institute of Technology/Deutsches Zentrum für Luft- und Raumfahrt) Prosthetic Hand II, an anthropomorphic controller is developed to help the amputees use and perceive the prosthetic hands more like people with normal physiological hands. The core of the anthropomorphic controller is a hierarchical control system. It is composed of a top controller and a low level controller. The top controller has been designed both to interpret the amputee's intensions through electromyography (EMG) signals recognition and to provide the subject-prosthesis interface control with electro-cutaneous sensory feedback (ESF), while the low level controller is responsible for grasp stability. The control strategies include the EMG control strategy, EMG and ESF closed loop control strategy, and voice control strategy. Through EMG signal recognition, 10 types of hand postures are recognized based on support vector machine (SVM). An anthropomorphic closed loop system is constructed to include the customer, sensory feedback system, EMG control system, and the prosthetic hand, so as to help the amputee perform a more successful EMG grasp. Experimental results suggest that the anthropomorphic controller can be used for multi-posture recognition, and that grasp with ESF is a cognitive dual process with visual and sensory feedback. This process while outperforming the visual feedback process provides the concept of grasp force magnitude during manipulation of objects.

Reactively dealing with self-collisions is an important requirement on multidegree-of-freedom robots in unstructured and dynamic environments. Classical methods to integrate respective algorithms into task hierarchies cause substantial problems: Either these unilateral safety constraints are permanently active, unnecessarily locking DOF for other tasks, or they get activated online and result in a discontinuous control law. We propose a new, reactive self-collision avoidance algorithm for highly complex robotic systems with a large number of DOF. In particular, configuration-dependent damping is imposed to dissipate undesired kinetic energy in a well-directed manner. Moreover, we merge the algorithm with a novel method to incorporate these unilateral constraints into a dynamic task hierarchy. Our approach both allows us to specifically limit the force/torque derivative to comply with physical constraints of the real robot and to prevent discontinuities in the control law while activating/deactivating the constraints. No redundancy is wasted. No comparable algorithms have been developed and implemented on a torque-controlled robot with such a level of complexity so far. The implementation of our generic solution on the multi-DOF humanoid Justin clearly validates the performance and demonstrates the real-time applicability of our synthetic approach. The proposed method can be used to contribute to whole-body controllers.

The mobile humanoid Rollin'Justin is a versatile experimental platform for research in manipulation tasks. Previously, different state of the art control methods and first autonomous task execution scenarios have been demonstrated. In this video two new applications with challenging task requirements are presented. One is the catching of one or even two flying balls using all of Justin's degrees of freedom. The other is the autonomous preparation of coffee. Both applications need adequate sensors to support local referencing. The required precision in position and timing is realized in software, using the sensor information, taking the varying precision of Justin's kinematic sub-chains into account and handling all timings in sub-millisecond range.

In recent years a new generation of motion simulators, based on serial kinematics industrial robots, emerged as alternative to the currently prevalent Steward-platforms. This paper presents the newest addition to this: The DLR Robot Motion Simulator. Part I covers the design process and gives a detailed introduction of the setup. The overall layout of the simulation platform and its parts is introduced. To meet the requirements of an interactive simulation, a new piloting cell had to be designed using light-weight construction and real time vehicle simulations were required. Another key issue of the whole design is the guarantee of safety. The simulator utilizes a 10m linear axis to increase its workspace to provide a better simulation experience compared to previous designs of motion simulators with serial or parallel kinematics. Part II introduces a new path planning algorithm for the kinematically redundant simulator platform, required for the generation of appropriate motion cues. Furthermore the application artistic-flight simulation is demonstrated and validated.

In Part I of this paper, a novel motion simulator platform is presented, the DLR Robot Motion Simulator with 7 degrees of freedom (DOF). In this Part II, a path-planning algorithm for mentioned platform will be discussed. By replacing the widely used hexapod kinematics by an antropomorhic, industrial robot arm mounted on a standard linear axis, a comparably larger workspace at lower hardware costs can be achieved. But the serial, redundant kinematics of the industrial robot system also introduces challenges for the path-planning as singularities in the workspace, varying movability of the system and the handling of robot system's kinematical redundancy. By solving an optimization problem with constraints in every sampling step, a feasible trajectory can be generated, fulfilling the task of motion cueing, while respecting the robot's dynamic constraints.

Despite the progress since the first attempts of mankind to explore space, it appears that sending man in space remains challenging. While robotic systems are not yet ready to replace human presence, they provide an excellent support for astronauts during maintenance and hazardous tasks. This paper presents the development of a space qualified multi-fingered robotic hand and highlights the most interesting challenges. The design concept, the mechanical structure, the electronics architecture and the control system are presented throughout this overview paper.

One characteristic attribute of mobile platforms equipped with a set of independent steering wheels is their omnidirectionality and the ability to realize complex translational and rotational trajectories. An accurate coordination of steering angle and spinning rate of each wheel is necessary for a consistent motion. Since the orientations of the wheels must align to the Instantaneous Center of Rotation (ICR), the current location and velocity of this specific point is essential for describing the state of the platform. However, singular configurations of the controlled system exist depending on the ICR, leading to unfeasible control inputs, i.e., infinite steering rates. Within this work we address and analyze this problem in general. Furthermore, we propose a solution for mobile platforms with variable footprint. An existing controller based on dynamic feedback linearization is augmented by a new potential field-based algorithm for singularity avoidance which uses the tunable leg lengths as an additional control input to minimize deviations from the nominal motion trajectory. Simulations and experimental results on the mobile platform of DLR's humanoid manipulator Justin support our approach.

One of the fundamental demands on robotic systems is a safe interaction with their environment. For fulfilling that condition, both collisions with obstacles and the own structure have to be avoided. We address the problem of self collisions and propose an algorithm for its avoidance which is based on artificial repulsion potential fields and applicable to both torque and position controlled manipulators. To this end, we design a damping that incorporates the configuration dependance of the robot. For a maximum level of safety, an additional emergency brake strategy based on kinetic energy considerations is introduced for situations in which self collisions are not avoidable by the controller. Experiments are performed on DLR's humanoid Justin.

An anthropomorphic hand arm system using variable stiffness actuation has been developed at DLR. It is aimed to reach its human archetype regarding size, weight and performance. The main focus of our development is put on robustness, dynamic performance and dexterity. Therefore, a paradigm change from impedance controlled, but mechanically stiff joints to robots using intrinsic variable compliance joints is carried out. Collisions of the rigid joint robot at high speeds with stiff objects induce the energy too fast for an active controller to prevent damages. In contrast, passively compliant robots are able to temporarily store energy. In this case the resulting internal forces applied to the robot structure and the drive trains are reduced. Furthermore, the energy storage allows to outperform the dynamics of stiff robots. The hand drives and the electronics are completely integrated within the forearm. Extremely miniaturized electronics have been developed to drive the 52 motors of the system and interface their sensors. Several variable stiffness actuation principles used in the arm joints and the hand are presented. The paper highlights the different requirements that they have to fulfill. A first test of the systems robustness and dynamics has been performed by driving nails with a grasped hammer and is demonstrated in the attached video.

The compensation of motion of the beating heart is investigated in the context of minimally invasive robotic surgery. Although reduced by mechanical stabilisers, residual tissue motion makes surgery still difficult and time consuming. Compensation for this motion is therefore highly desirable. Motion can be captured by tracking natural landmarks on the heart surface viewed by a video endoscope. The proposed motion compensation scenario uses tracked motion information on the beating heart to stabilise image motion at the surgeon's display and to move his instruments accordingly, i.e. synchronously to the surface of the beating heart. The setup uses a specially designed heart motion simulator to display real heart motion and employs a particular prediction strategy to overcome latencies, which is required to achieve synchronous motion of robot instruments and the heart surface. Results show that the proposed methods are suitable to achieve the desired compensation of heart motion.

This video presents the complete procedure for the optimal setup of the DLR MiroSurge telerobotic system for minimally invasive surgery. Two key features are implemented. First, optimization algorithms preoperatively determine several setups that are then rated and selected by the surgeon. Second, the intraoperative situation is taken into account. The newly developed VR-Map device together with fast registration and optimization algorithms enable a quick procedure to assure the optimal patient-specific setup of the robotic system.

The procedure of manually generating a 3D model of an object is very time consuming for a human operator. Next best-view (NBV) planning is an important aspect for automation of this procedure in a robotic environment. We propose a surface-based NBV approach, which creates a triangle surface from a real-time data stream and determines viewpoints similar to human intuition. Thereby, the boundaries in the surface are detected and a quadratic patch for each boundary is estimated. Then several viewpoint candidates are calculated, which look perpendicular to the surface and overlap with previous sensor data. A NBV is selected with the goal to fill areas which are occluded. This approach focuses on the completion of a 3D model of an unknown object. Thereby, the search space for the viewpoints is not restricted to a cylinder or sphere. Our NBV determination proves to be very fast, and is evaluated in an experiment on test objects, applying an industrial robot and a laser range scanner.

Many real-world tasks require fast planning of highly dynamic movements for their execution in real-time. The success often hinges on quickly finding one of the few plans that can achieve the task at all. A further challenge is to quickly find a plan which optimizes a desired cost. In this paper, we will discuss this problem in the context of catching small flying targets efficiently. This can be formulated as a non-linear optimization problem where the desired trajectory is encoded by an adequate parametric representation. The optimizer generates an energy-optimal trajectory by efficiently using the robot kinematic redundancy while taking into account maximal joint motion, collision avoidance and local minima. To enable the resulting method to work in real-time, examples of the global planner are generalized using nearest neighbour approaches, Support Vector Machines and Gaussian process regression, which are compared in this context. Evaluations indicate that the presented method is highly efficient in complex tasks such as ball-catching.

This video presents our recent research on the integration of physical human-robot interaction (pHRI) into imitation learning. First, a marker control approach for real time human motion imitation is shown. Secondly, physical coaching in addition to observational learning is applied for the incremental learning of motion primitives. Last, we extend imitation learning to learning pHRI which includes the establishment of intended physical contacts. The proposed methods were implemented and tested using the IRT humanoid robot and DLR's humanoid upper-body robot Justin.

Independent Contact Regions allow a robust finger placement on the object, despite of potential errors in finger position. They are computed without considering the kinematics of the end-effector, and are usually applied to off-line grasp planners. This paper presents an approach to obtain Reachable Independent Contact Regions by including the hand kinematics in the computational loop. The regions are computed in a short time, which allows real-time applications in virtual grasping. Potential applications of the proposed approach include regrasp planning, and dual-hand manipulation of objects.

This work aims at accurate estimation of the pose of a close-range 3-D modeling device in real-time, at high-rate, and solely from its own images. In doing so, we replace external positioning systems that constrain the system in size, mobility, accuracy, and cost. At close range, accurate pose tracking from image features is hard because feature projections do not only drift in the face of rotation but also in the face of translation. Large, unknown feature drifts may impede real-time feature tracking and subsequent pose estimation-especially with concurrent operation of other 3-D sensors on the same computer. The problem is solved in Ref. [1] by the partial integration of readings from a backing inertial measurement unit (IMU). In this work we avoid using an IMU by improved feature matching: full utilization of the current state estimation (including structure) during feature matching enables decisive modifications of the matching parameters for more efficient tracking-we hereby follow the Active Matching paradigm.

This paper introduces an ultra-tightly coupled approach to data fusion of optical and inertial measurements. The two redundant sensor systems complement each other well, with the cameras providing absolute positions and the inertial measurements giving low latency information of derivatives. The targeted application is the tracking of handheld input devices for robotic surgery, where landmarks are not always visible to all cameras. Especially when bi-manual operation is considered, where one hand can move between the other hand and a camera, occlusions occur frequently. The ultra-tighly coupled data fusion uses 2D-camera measurements to correct pose estimations in an extended Kalman filter without an explicit 3D-reconstruction. Therefore marker measurements are used to support the pose estimation, even if the marker is only visible in one camera. Experiments were done with an inertial measurement unit and rectified stereo cameras that show the advantage of the approach for the application.

Bringing mechanically compliant joints to robots is in the focus of interest world wide, especially in the humanoid robotics community. Variable Stiffness Joints (VSJ) promise to gain a high performing and robust robotic system. The presented DLR Floating Spring Joint (FSJ) is a VSJ module designed for the first 4 axes of the anthropomorphic DLR Hand Arm System. The DLR Hand Arm System aims to match the skills of its natural archetype. For this purpose, the joints have to be extremely compact to fit into the arm. At the same time they require a high power density in order to approximate the human arm skills. The new DLR FSJ is designed completely from an energy based point of view. This addresses not only energy efficient components and low friction design, but also that the potential energy of the spring is used as good as possible. A demonstration of robustness is given by the investigation of a blunt impact to the tip of the arm.

It contradicts the human's expectations when humanoid robots move awkwardly during manipulation tasks. The unnatural motion may be caused by awkward start or goal configurations or by probabilistic path planning processes that are often used. This paper shows that the choice of an arm's target configuration strongly effects planning time and how human-like a planned path appears. Human-like goal configurations are found using a criterion from ergonomics research. The knowledge which pose of the Tool Center Point (TCP) can be reached in a natural manner is encapsulated in a restricted reachability map for the robot arm.

In this paper, a decentralized platform for Simultaneous Localization and Mapping (SLAM) with multiple robots is developed. A novel occupancy grid map fusion algorithm is proposed. Map fusion is achieved through a multi-step process that includes image pre-processing, map learning, relative transformation extraction and then verification of the results. The proposed map learning method is a process based on the Self Organizing Map (SOM). In the learning phase, the obstacles of the map are learned by clustering the occupied cells of the map. The clusters represent the spatial form of the map and make further analyses of the map easier and faster. Also, clusters can be interpreted as features extracted from the occupancy grid map so the map fusion problem becomes a task of matching features. Results of the experiments from tests performed on a real environment with multiple robots prove the effectiveness of the proposed solution.

This paper builds up on the Capture Point concept and exploits the simple form of the dynamical equations of the Linear Inverted Pendulum model when formulated in terms of the center of mass and the Capture Point. The presented methods include (i) the derivation of a Capture Point (CP) control principle based on the natural dynamics of the linear inverted pendulum (LIP), which stabilizes the walking robot and motivates (ii) the design of a CP tracking and a CP end-of-step controller. The exponential stability of the CP control law is proven. Tilting is avoided by proper projection of the commanded zero moment point. The robustness of the derived control algorithms is analyzed analytically and verified in simulation and experiments.

The DLR Hand Arm System is based upon the variable stiffness concept which has been recently developed to improve impact robustness and energy efficiency of modern robots. This paper continues the work on the bidirectional antagonistic variable stiffness (BAVS) joint concept which is an extension of antagonistic joints. Three mechanical setups utilizing different spring and cam disc combinations to implement a desired torque-stiffness characteristic are analyzed. Two BAVS joint solutions as used for the wrist and forearm rotation of the DLR Hand Arm System are presented. Furthermore in the experimental section torque-deflection calibration and drive redundancy are validated.

A robotic heart motion simulator (HMS) is developed at DLR (German Aerospace Center) to accurately simulate real translational motions of a mechanically stabilized beating heart in a lab environment. This simulator is part of the DLR scenario for motion compensation on the beating heart. Motion compensation is a highly desired issue in minimally invasive surgery. The design of the HMS as well as its derivation of inverse kinematics, closed-loop dynamics, and proposed control structures are presented. The simulator is evaluated in the lab environment to prove its capability to precisely simulate low-amplitude translational motions of a stabilized beating heart.

The computing and communication architecture of the DLR Hand Arm System is presented. Its task is to operate the robot's 52 motors and 430 sensors. Despite that complexity, the main design goal for it is to create a flexible architecture that enables high-performance feedback control with cycles beyond 1kHz. Flexibility is achieved through a hierarchical net of computing nodes that goes from commercial-of-the-shelf hosts down to the physical interfaces of sensors and actuators. The concept of a Hardware Abstraction Layer (HAL) provides a convenient high-level interface to the entire robotic hardware. First experiments with prototypical control applications, featuring 100 kHz and 3 kHz control loops, demonstrate the performance of the architecture.

This paper presents a method to calibrate the model of serial and flexible lightweight robots with joint sided torque sensors in the assembled state. The calibration is done in an iterative three-step process, based on static robot poses. In the first step the kinematics and stiffnesses of the flexible components are calibrated. Second the models of the integrated torque sensors are identified in a linear least square solution. In the third step the masses, the centers of gravity and the torque sensor offsets are estimated using linear regression. The calibration steps are repeated stepwise to account for their dependencies. The calibration procedure is simulated and experimentally performed with the medical lightweight robot MIRO of the German Aerospace Center. Through the iterative procedure the pose accuracy improves from about 5mm translational error and 2.5 ° rotational error to 1mm and 0.3 ° regarding the entire workspace.

This paper presents the graspability map, a novel approach to represent for a particular object the positions and orientations that a given mechanical hand can adopt to achieve a force closure precision grasp. The algorithm is based on the intersection between the fingertip workspaces and the object, plus the verification of a necessary condition for force closure grasps. The maps are computed offline and can be used for comparing the grasp capabilities of different mechanical hands with respect to some benchmark objects. The maps have also potential applications in online grasp and manipulation planning.

This paper introduces the mechatronic design and a first performance analysis of a new haptic exoskeleton, the X-Arm-2. The X-Arm-2 is a fully actuated force-reflecting human arm exoskeleton that is based on our previously proposed approach to ergonomic and human-centered exoskeleton design [1] [2]. The X-Arm-2 is a highly power-dense impedance-type haptic device that (1) can interact with natural human arm movement of varying operators without requiring adjustments and creating constraint forces, (2) provides crisp force-feedback through high actuator bandwidth, low residual friction and good joint torque sensor resolution, (3) has a low total mass of only 6.2 kg and (4) low inertia through a human-oriented mixed implementation of Bowden-cable relocated and directly-integrated DC actuators.

This paper presents the design and control of the sigma.7 haptic device and the new surgical console of the MiroSurge robotic system. The console and the haptic devices are designed with respect to requirements in minimally invasive robotic surgery. Dedicated left and right handed devices are integrated in an operator console in an ergonomic configuration. The height of the whole console is adjustable, allowing the surgeon seated and standed operation. Each of the devices is fully actuated in seven degrees of freedom (DoF). A parallel mechanism with 3 DoF actuates the translational motion and an attached wrist with 3 intersecting axis drives the rotations of the grasping unit. This advantageous design leads to inherently decoupled kinematics and dynamics. Cartesian forces are 20 N within the translational workspace, which is a sphere of about 120 mm diameter for each device. The rotational wrist of the device covers the whole workspace of the human hand and provides maximum torques of about 0.4 Nm. The grasping unit can display forces up to 8 N. An integrated force/torque sensor is used to increase the transparency of the devices by reducing inertia and friction. It is theoretically shown that the non-linear closed loop system behaves like a passive system and experimental results validate the approach. The sigma.7 haptic devices are designed by Force Dimension in cooperation with the German Aerospace Center (DLR). DLR designed the surgical console and integrated the haptic devices in the MiroSurge system.

In many robotic manipulation tasks a robotic hand is used to just grasp and fix the object of interest while the object motion is performed by the arm. Motivated by the analysis on human grasping using data reduction techniques, we applied this concept to the DLR Hand II. Therefore, we analyzed the grasp database that was grown over the past years to find suitable robotic ¿synergy coordinates¿. 74% of these grasps can be represented by two coordinates that were originally defined by 12 joint variables. As a second step, a synergy impedance controller was derived and implemented extending the work on passivity based hand control at DLR. This controller for torque-controlled robot hands allows to imitate the behavior of a synergistic, respectively underactuated, hand. Such a controller provides furthermore a simplified interface for higher level grasping strategies and allows furthermore to manually teach new grasps easily. The controller was evaluated on the DLR Hand II by commanding steps that demonstrate the desired transient behavior. Finally, two objects were successfully grasped validating our approach.

The collaborative research center SFB453 (www.sfb453.de) aims to realize high-fidelity telepresence and teleaction systems. Telepresence and teleaction systems extend the human workspace to remote locations in order to overcome barriers like distance, scaling, danger or the human skin. Using a human-system interface the human operator controls a remotely located teleoperator. Multi-modal feedback in form of visual, auditory, and haptic data is used to increase the feeling of telepresence. Different application areas including minimally invasive surgery, on-orbit servicing, microassembly as well as tele-manufacturing and tele-maintenance are targeted.

Tendon driven mechanisms have been considered in robotic design for several decades. They provide lightweight end effectors with high dynamics. Using remote actuators it is possible to free more space for mechanics or electronics. Nevertheless, lightweight mechanism are fragile and unfortunately their control software can not protect them during the very first instant of an impact. Compliant mechanisms address this issue, providing a mechanical low pass filter, increasing the time available before the controller reacts. Using adjustable stiffness elements and an antagonistic architecture, the joint stiffness can be adjusted by variation of the tendon pre-tension. In this paper, the fundamental equations of m antagonistic tendon driven mechanisms are reviewed. Due to limited tendon forces the maximum torque and the maximum acheivable stiffness are dependent. This implies, that not only the torque workspace, or the stiffness workspace must be considered but also their interactions. Since the results are of high dimensionality, quality measures are necessary to provide a synthetic view. Two quality measures, similar to those used in grasp planning, are presented. They both provide the designer with a more precise insight into the mechanism.

In this paper we introduce a classification of intrinsically compliant joint mechanisms. Furthermore, we outline design considerations for realizing such devices in order to match the requirements for robust and performant actuation. Based on this elaboration, a new design concept is presented, the DLR QA-Joint. Its performance is investigated by various experiments, covering velocity increase using the elastic energy, joint protection capabilities, and control performance.

This article analyzes and evaluates the stability of the biologically inspired gait of the DLR Crawler, a walking hexapod robot, with respect to leg loss. Using a kinematic simulation, ranges of velocity commands that result in stable gait coordination are determined for both cases, the undamaged robot and the robot experiencing the loss of a single leg. The results give insight how to adjust the motion commands after the loss of a leg. Further, a simplified dynamic simulation is used to analyze the effect of leg loss on the walking stability. Heuristic measures like curvature and length of the traveled path, roll and pitch angles are employed to evaluate the walking stability and performance. Some methods like shifting the COG or stiffening the variably compliant joints are proposed and discussed with respect to their ability to improve the walking performance in case of leg loss. In the end, the presented concepts are extended and for the first time applied to a simulated eight-legged robot.

Up to now, mostly blunt human-robot impacts were investigated in the robotics literature. In this context, the influence of robot mass and velocity during rigid impacts with and without the possibility of the human being clamped was quantified. In this paper an analysis of soft-tissue injuries caused by sharp tools, which are mounted on/grasped by a robot is carried out as the next step down the road to a full safety analysis of robots for HRI. We conducted an analysis of soft-tissue injuries based on available biomechanical and forensic data and to our knowledge for the first time in robotics present various experimental results with biological tissue for validation. Furthermore, possible countermeasures are evaluated quantitatively based biomechanically relevant quantities.

This paper generalizes the Time Domain Passivity Control concept originally introduced by J.-H. Ryu et al. (2004) in order to work for multi-degree of freedom (DoF) haptic systems with time delay. Its energy computation (named passivity observer) factors in the phase shift caused by time delay, and is improved by an energy estimation. Moreover, the variable damping of the passivity controller is generalized such that weighting by the mass matrix of the haptic device is possible. This transformation takes into account the direction-dependent inertia of multi-DoF haptic devices. Furthermore, a stability boundary for this damping is introduced for one as well as for several DoF allowing for high energy dissipation. Additionally, it is briefly shown that one single multi-DoF Cartesian passivity controller is advantageous compared to independent single-DoF passivity controllers in each joint of the haptic device. Finally, the generalized Time Domain Passivity Controller is experimentally verified using the DLR light weight robot arm as haptic device.

Typical industrial assembly tasks require an accuracy that cannot be realized by only feedback control if a minimum speed is given by a conveyor. Feed-forward has proven to be advantageous, using predictions of the desired trajectory which will be computed from sensor values. These predictions are improved by a model based classification of the sensor data to typical scenarios. In contrast to linear controllers this assures the fastest possible response to external disturbances, in spite of large dynamical delays. The method is demonstrated by assembling wheels to a car body that is moved by a conveyor, fusing sensor data using an extended Kalman filter.

Within assembly lines, wheel assembly to continuously conveyed car bodies is still executed by human workers using a device that compensates the weight of the wheel. This paper presents a solution in which a robot autonomously assembles and fixes the wheels. The approach uses a sensor-driven control strategy that compensates a possible temporal or spatial offset. Three types of sensors are proposed for adequate perception of the wheel hub. Their signals are fused by a Kalman filter that allows predictions in the time domain. Finally, a feed-forward controller is used, that is designed to consider the predictions in order to minimize dynamical delays. The control is driven by a special task description that extents usual robot programming methods.

The variable stiffness actuation concept is considered to provide a human-friendly robot technology. This paper examines a joint concept called the bidirectional antagonistic joint which is a extension of antagonistic joints. A new operating mode called the helping mode is introduced, which increases the joint load range. Although the joint can not be pretensioned in the helping mode, it is shown that a stiffness variation is possible, assuming a suitable torque-stiffness characteristic of the elastic elements. A methodology to design such characteristics is presented along with several example cases interpreted in a torque-stiffness plot. Furthermore, a stiffness adaptation control scheme which ensures mechanism safety is described. Finally, the design methodology and the control are evaluated on an implementation of a bidirectional antagonistic joint.

This paper proposes a method for accurate robotic motion compensation of a freely moving target object. This approaches a typical problem in medical scenarios, where a robotic system needs to compensate physiological movements of a target region related to the patient. An optical tracking system measures the poses of the robot's end-effector and the moving target. The task is to track the target with the robot in a desired relative pose. Arbitrary motion in 6 DoF is covered. The position controller of the medical light-weight robot MIRO is enhanced by a Cartesian displacement observer. The proposed observer feedback preserves the dynamics of the robot, while achieving high accuracy in task space. The target object is equipped with an inertial measurement unit in addition to tracking markers. Target sensor data is fused by an extended Kalman filter in a tightly coupled approach. The robot control and the target tracking in the task space aim to combine accuracy, dynamic performance and robustness to marker occlusions. The algorithms are verified with the DLR MIRO, an experimental target platform, and a commercial tracking system. The experiments demonstrate rapid convergence to desired Cartesian poses and good dynamic tracking performance even at higher target motion speed.

A robotic ball-catching system built from a multi-purpose 7-DOF lightweight arm (DLR-LWR-III) and a 12 DOF four-fingered hand (DLR-Hand-II) is presented. Other than in previous work a mechatronically complex dexterous hand is used for grasping the ball and the decision of where, when and how to catch the ball, while obeying joint, speed and work cell limits, is formulated as an unified nonlinear optimization problem with nonlinear constraints. Three different objective functions are implemented, leading to significantly different robot movements. The high computational demands of an online realtime optimization are met by parallel computation on distributed computing resources (a cluster with 32 CPU cores). The system achieves a catch rate of > 80% and is regularly shown as a live demo at our institute.

In this paper we present the synchronization and driver architecture of the DLR LWR-III, which supplies an easy to use interface for applications. For our purpose we abstracted the robot hardware entirely from the control algorithms using the common device driver concept of modern operating systems. The software architecture is split into two modular parts. On the one side, there are device drivers that communicate with the hardware components. On the other side, there are realtime applications realized as Simulink Models, which provide advanced control algorithms. This ensures a clean separation between the two modules and provides a communication over a common and approved interface. Furthermore we investigated how we can ensure synchronization to the hardware over the device driver interfaces and how we can ensure that it meets hard realtime requirements. The main result of this paper is to realize a synchronization between LWR-III hardware and Simulink control applications while targeting small latencies with respect to hard realtime requirements. The design is implemented and verified on WindRiver™VxWorks™.

The impressive manipulation capabilities of the human hand are undoubtedly related to the thumb opposition. Such a versatility is highly desirable in the context of humanoid robots, in particular when performing object manipulation. Biomechanical data, surgery procedures and rehabilitation surveys represent an excellent base from which a robotic design can be inferred. This knowledge must be understood to identify the properties required for manipulation skills, and especially, to obtain a holistic view of the thumb functionality. Several designs have been realized, that concentrated on biomimetism or on classical mechanism designs. Therefore, it is currently difficult for designers to obtain a clear overview of the properties required for a functional robot thumb. In the present case, a robotic hand with size, forces, velocity and shape comparable to the human ones, is envisioned. Unlike most of robotic designs-where the fingers are modular and the thumb is simply a finger placed in opposition-the thumb benefits from an intensive functional analysis. This paper gathers anatomy, surgery and rehabilitation data and identifies the properties required for human like manipulation. Based on this synergy, guidelines are presented that are fused and applied to the hand design of the Integrated Hand arm project of DLR.

This paper presents a semi-automatic robotic system supporting a surgeon in the harvesting of the internal mammary artery (IMA) for an open chested intervention in coronary revascularisation surgery. The versatile surgical lightweight robot MIRO developed at DLR (German Aerospace Center) is used to detect and mark the path of the IMA at the inner side of the thoracic wall. The robot is equipped with a tool combining a Doppler ultrasonography (US) probe and a medical marker pen. The position of the IMA is extracted from the US-images to place the tool above the artery via visual servoing. Additionally, the robot moves the tool in direction of the artery to mark the location of the IMA on it's path. To achieve an ideal contact situation for US-imaging along the whole path the contact force between tissue and probe is controlled according to force measurements based on the internal torque sensors of the robot. The evaluation of the robotic system by an animal experiment shows that the system is capable of robustly detecting the IMA.

In this paper we present various new insights on the effect intrinsic joint elasticity has on safety in pHRI. We address the fact that the intrinsic safety of elastic mechanisms has been discussed rather one sided in favor of this new designs and intend to give a more differentiated view on the problem. An important result is that intrinsic joint elasticity does not reduce the Head Injury Criterion or impact forces compared to conventional actuation with some considerable elastic behavior in the joint, if considering full scale robots. We also elaborate conditions under which intrinsically compliant actuation is potentially more dangerous than rigid one. Furthermore, we present collision detection and reaction schemes for such mechanisms and verify their effectiveness experimentally.

In this overview paper we present current work on safety analysis for physical Human-Robot Interaction (pHRI) and motion control methods for robotic co-workers. In particular, we introduce the analysis tools for investigating the potential injury a human would suffer during robot-human impacts. Furthermore, we outline our concept for establishing a procedure towards standardized crash testing in robotics with automobile crash-test dummies. Since it is only possible to investigate blunt impacts with these devices, we developed a drop testing setup for analyzing soft-tissue injury in robotics from a biomechanics perspective. In the second part of the paper, some of our methods for task preserving and task relaxing motion schemes are described, which enable collision avoidance in real-time. The algorithms are well suited to work in an integrated fashion with the soft robotics control developed for the DLR Lightweight Robot III (LWR-III). In addition, it is shown how the torque sensing capabilities of the robot can be used to support reactive motion schemes. Finally, an overview of our human-friendly control architecture for the LWR-III is given, which unifies the rich bundle of developed methods for this manipulator.

This paper describes a novel method for motion generation and reactive collision avoidance. The algorithm performs arbitrary desired velocity profiles in absence of external disturbances and reacts if virtual or physical contact is made in a unified fashion with a clear physically interpretable behavior. The method uses physical analogies for defining attractor dynamics in order to generate smooth paths even in presence of virtual and physical objects. The proposed algorithm can, due to its low complexity, run in the inner most control loop of the robot, which is absolutely crucial for safe Human Robot Interaction. The method is thought as the locally reactive real-time motion generator connecting control, collision detection and reaction, and global path planning.

This paper introduces a study aimed to help quantify the benefits of limited-performance force-feedback user input devices for space telemanipulation with a dexterous robotic arm. A teleoperated robotic hand has been developed for the European Space Agency by the German Aerospace Center (DLR) for a lunar rover prototype. Studies carried out on this telerobotic system investigated several criteria critical to telemanipulation in space: (1) grasping task completion time, (2) grasping task difficulty, (3) grasp quality, and (4) difficulty level for the operator to assess the grasp quality. Several test subjects were allocated to remotely grasp regular and irregular shaped objects, under different combinations of visual- and force-feedback conditions. This work categorized the benefits of visual- and force-feedback in teleoperated grasping through several performance metrics. Furthermore, it has been shown that, with local joint-level impedance control, good grasping performance with rigid hard objects can be achieved, even with limited force-feedback information and low communication bandwidth. On the other hand, a performance ceiling was also found when grasping deformable objects, where the limited force-feedback setup cannot sufficiently reflect the object boundary to the teleoperator.

Robotic surgery systems are highly complex and expensive pieces of equipment. Demands for lower cost of care can be met if these systems are employable in a flexible manner for a large variety of procedures. To protect the initial investment the capabilities of a robotic system need to be expandable as new tasks arise. To oblige the needs of future robotic support in hospitals, the German Aerospace Center (DLR) has developed the versatile robotic system MiroSurge for medical applications. This paper presents a 3 DoF instrument for Minimally Invasive Robotic Surgery which is mounted to the hollow wrist of the DLR MIRO robot arm. The MICA instrument consists of a versatile drive train and a detachable task specific tool with its tool interface, shaft, 2 DoF wrist, 7 DoF force/torque sensor and the actuated functional end. With the current cable-driven tool, gripping and manipulation forces of above 10 N are feasible and dynamics is high enough for surgery at the beating heart.

In virtual assembly verification or remote maintenance tasks, bimanual haptic interfaces play a crucial role in successful task completion. This paper proposes a method for objectively comparing how well a haptic interface covers the reachable workspace of human arms. Two system configurations are analyzed for a recently introduced haptic device that is based on two DLR-KUKA light weight robots: the standard configuration, where the device is opposite the human operator, and the ergonomic configuration, where the haptic device is mounted behind the human operator. The human operator directly controls the robotic arms using handles. The analysis is performed using a representation of the robot arm workspace. The merits of restricting the comparisons to the most significant regions of the human workspace are discussed. Using this method, a greater workspace correspondence for the ergonomic configuration was shown.

In autonomous bimanual operation of a robot, parallelized planning and execution of a task is essential. Elements of a task have different functional and spatial relationships. They may depend on each other and have to be executed in a specific order or they may be independent and their order can be determined freely. Consequently, individual actions can be planned and executed in parallel or not. In a proof of concept, this paper shows that the structure of a task and its mapping onto subordinate planners can significantly influence planning speed and task execution. Independent tasks are planned using two parallel path planners. Dependent tasks are planned using one path planner for both arms. Using a simple, yet expandable experimentation scenario, the resulting recommendations for parameterizing path planners are verified on a humanoid robot. For execution on the real robot a violation of the rigid body model used in path planners had to be addressed.

Research on humanoid robots for use in servicing tasks, e.g. fetching and delivery, attracts steadily more interest. With “Rollin' Justin” a mobile robotic system and research platform is presented that allows sophisticated control algorithms and dexterous manipulation. This video gives an overview of the mobile humanoid robotic system “Rollin' Justin” with special emphasis on mechanical design features, control issues and high-level system capabilities such as human robot interaction.

Research on humanoid robots for use in servicing tasks, e.g. fetching and delivery, attracts steadily more interest. With Rollin' Justin a mobile robotic system and research platform is presented that allows the implementation and demonstration of sophisticated control algorithms and dexterous manipulation. Important problems of service robotics such as mobile manipulation and strategies for using the increased workspace and redundancy in manipulation task can be studied in detail. This paper gives an overview of the design considerations for a mobile platform and their realizations to transform the formerly table-mounted humanoid upper body system Justin into Rollin' Justin, a fully self-sustaining mobile research platform.

Mobile platforms equipped with several steering wheels are known to be omnidirectional, i.e., able to independently translate and rotate on the plane. As an improvement to this design, the Justin mobile platform also possesses the ability to vary its footprint over time by extending/retracting the wheel legs during motion. In this paper, we discuss the kinematic modeling and control issues for such a platform. The goal is to obtain a tracking controller which is able to realize an arbitrary linear/angular platform motion while, at the same time, independently expanding/retracting each leg. Experimental results support the proposed approach.

After giving a rich data basis of our impact tests with standardized crash-test dummies in Part I of this work we address in Part II various aspects related to these tests in a case based discussion. The presented facts, the knowledge gained from our previous work, and the data from Part I lead us to recommendations for standardized crash-testing procedures in robotics. The proposed impact procedures will help to compare blunt robot-human impacts on a common basis. We will discuss additional requirements which will enhance the completeness of testing procedures.

After analyzing fundamental impact characteristics of robot-human collisions in our previous work, the intention in the present paper is to augment existing knowledge in this field, verify previously given statements with standardized equipment of the German Automobile Club (ADAC), and provide a crash-test report for robots in general. Various new insights are achieved and a systematic and extensive set of data is provided. The presented work is divided into two papers. The main purpose of Part I is to give, similarly to reports known from the automobile world1, a fact based and result oriented view on our newest robot crash-test experiments. In Part II detailed discussions of the results listed in the present paper and recommendations towards a standard crash-test protocol for robot safety are carried out.

This paper presents inverse position kinematics algorithms with real time capability for Justin, a robotic system with high redundancy and many degrees of freedom. The combination of closed form solutions for parts of the kinematic chain embedded in a nonlinear equation solver is shown to be advantageous. The algorithms are evaluated with DLR's robot Justin both in simulation and reality. Calculation times of 1 ms are achieved, including various optimization criteria for redundancy resolution. In case only a single arm with 7 DoF is considered, a fast calculation time of 250 µs is reached. With inclusion of an iterative step, reachability can be shown in more than 99% of the calculations regardless of the initial guess. The problem of weighting in multi-criteria optimization problems remains, though in the chosen approach the tool tip position is never compromised by other criteria due to the partially closed form solution. The presented algorithm can be applied to inverse position kinematics for all manipulators with serial or tree structure and redundant joints in case closed form solutions are available for parts of the kinematic chain.

This video presents the in-house developed DLR MiroSurge robotic system for surgery. As shown, the system is suitable for both minimally invasive and open surgery. Essential part of the system is the MIRO robot: The soft robotics feature enables intuitive interaction with the robot.

A new lossy compression method is proposed for haptic (force, velocity) data as exchanged in bilateral telepresence systems. The method is based on the passive extrapolative compression strategy proposed in the work of Kuschel et al. (2006). The innovation is that the extrapolations do not have a stiff horizon, but are triggered by considerable changes (events) in the target environment. This enables longer average extrapolation horizons and thus, higher compression. Experiments are conducted using two DLR light weight robots. The results indicate that the method outperforms older implementations.

This video presents a telepresence system which enables a human operator to explore a remote environment by means of a multimodal man machine interface and rollin' JUSTIN as teleoperator. The man machine interface allows for bimanual, dexterous manipulation and, through two different operating modi of the man machine interface, wide area movement as well. A bimanual assembly task, consisting of grasping a connector, opening and closing it again, is shown in this video.

This paper introduces the planning and control software of a teleoperation system for research in minimally invasive robotic surgery. It addresses the problem of how to organize a complex system with 41 degrees of freedom as a flexible configurable platform. Robot setup planning, force feedback control and nullspace handling with three robotic arms are considered. The planning software is separated into sequentially executed planning and registration procedures. An optimal setup is first planned in virtual reality and then adapted to variations in the operating room. The real time control system is structured in hierarchical layers. Functions are arranged in the layers with respect to their domain and maximum response time. The design is flexible and expandable while performance is maintained. Structure, functionality and implementation of planning and control are described. The prototypic robotic system provides intuitive bimanual bilateral teleoperation within the planned working space.

This paper proposes an adaptive controller for a fully free-floating space robot with kinematic and dynamic model uncertainty. In adaptive control design for the space robot, because of high dynamical coupling between an actively operated arm and a passively moving end-point, two inherent difficulties exist, such as non-linear parameterization of the dynamic equation and both kinematic and dynamic parameter uncertainties in the coordinate mapping from Cartesian space to joint space. The proposed method in this study overcomes the above two issues by paying attention to the coupling dynamics. The proposed adaptive controller does not involve any measurement of acceleration; but it is still possible for the system to be linearly parameterized in terms of uncertain parameters and a suitable input torque can be generated in the presence of model uncertainty. A numerical simulation was carried out to confirm the validity of the proposed adaptive control.

In the context of 3-D scene modeling, this work aims at the accurate estimation of the pose of a close-range 3-D modeling device, in real-time and passively from its own images. This novel development makes it possible to abandon using inconvenient, expensive external positioning systems. The approach comprises an ego-motion algorithm tracking natural, distinctive features, concurrently with customary 3-D modeling of the scene. The use of stereo vision, an inertial measurement unit, and robust cost functions for pose estimation further increases performance. Demonstrations and abundant video material validate the approach.

This paper considers the issue of calibrating a camera with narrow angular field of view using standard, perspective methods in computer vision. In doing so, the significance of perspective distortion both for camera calibration and for pose estimation is revealed. Since narrow angular field of view cameras make it difficult to obtain rich images in terms of perspectivity, the accuracy of the calibration results is expectedly low. From this, we propose an alternative method that compensates for this loss by utilizing the pose readings of a robotic manipulator. It facilitates accurate pose estimation by nonlinear optimization, minimizing reprojection errors and errors in the manipulator transformations at the same time. Accurate pose estimation in turn enables accurate parametrization of a perspective camera.

In this paper a sensor fusion for pose estimation using optical and inertial data is presented. The proposed algorithm is based on extended Kalman filtering and fuses data from an optical tracking system and an inertial measurement unit. These two redundant sensor systems complement each other well, with the tracking system providing absolute positions and the inertial measurements giving low latency information of derivatives. Models for both sensors are given respecting the different sampling times and latencies. Another key issue is to use information about every landmark, i.e. marker ball, visible for the tracking system, by coupling the two sensor systems tightly together. The algorithm is evaluated in simulation and tested with an experimental hardware platform. The combined sensor system is robust with respect to short time marker occlusions and effectively compensates for latencies in the pose measurements.

The capabilities of DLR's multi-DOF humanoid robot Justin are extended with the help of a dynamic torque control component for base reaction minimization. Since the mobile base of the robot comprises springs, reactions induced by arm/torso motions lead to vibrations and deteriorate the performance. The control component is derived from the equation of motion of the robot, represented as an underactuated system, and partitioned into a ¿driven¿ subsystem (one of the arms), and a ¿compensating¿ subsystem (the other arm, with or w/o torso contribution). The control component is then embedded into the existing sophisticated controller structure of Justin, as a feedforward component, with additional control signals from an augmented PD feedback controller. It was possible to obtain satisfactory performance with a very ¿soft¿ compensatory subsystem. The experimental results confirmed the potential of this model-based approach for use in a complex multi-DOF system. As far as we know, this is the first time that a dynamic-coupling compensating controller is applied to a real system of such complexity, utilizing thereby a torque control interface.

Telesurgery systems integrate multimodal communication and robotic technologies to enable surgical procedures to be performed from remote locations. They allow human surgeons to intuitively control laparoscopic instruments and to navigate within the human body. In this paper, we present selected topics on multimodal interaction in the context of telesurgery applications. These are results from the collaborative research project SFB 453 on ldquoHigh-Fidelity Telepresence and Teleactionrdquo which is funded by the German Research Foundation in the larger Munich area. The focus in this paper is on multimodal information processing and communication including simulation of surgical targets in the human body. Furthermore, we present an overview of our advanced multimodal telesurgery demonstrators that provide a comprehensive platform for our collaborative telepresence research.

Recently, ToF-cameras have attracted attention because of their ability to generate a full 2 1/2D depth image at video frame rates. Thus, ToF-cameras are suitable for real-time 3D tasks such as tracking, visual servoing or object pose estimation. The usability of such systems mainly depends on an accurate camera calibration. In this work a calibration process for ToF-cameras with respect to the intrinsic parameters, the depth measurement distortion and the pose of the camera relative to a robotpsilas end effector is described. The calibration process is not only based on the monochromatic images of the camera but also uses its depth values that are generated from a chequer-board pattern. The robustness and precision of the presented method is assessed applying it to randomly selected shots and comparing the calibrated measurements to a ground truth obtained from a laser scanner.

The time domain passivity framework is attracting interest as a method for granting stability in both telerobotics and haptic contexts; this paper employs this approach in order to introduce a novel concept, the Bilateral Energy Transfer for haptic telepresence. Loosely speaking, the Bilateral Energy Transfer is the straightforward transfer of energy between the two opposite sides of a teleoperation network, the master and slave robots. In an ideal telepresence scenario master and slave robots behave as rigid connected masses [1], and their power exchange is lossless; conversely, realistic scenarios include sources of energy leaks, i.e. elements that modify the power flows in the network. Moreover, if energy leaks have an active nature, they become source of instability for the system. This work isolates two sources of instability normally present in a teleoperation system, i.e. the delayed communication channel and robot velocity estimation based on digital position acquisition. These energy leaks are counterbalanced by two independent controllers, whose design is based on energetic consideration, and whose employment allows to achieve the Bilateral Energy Transfer. The presented arguments are sustained by simulations and experiments.

The dexterity of active hand prosthetics is limited not only due to the limited availability of dexterous prosthetic hands, but mainly due to limitations in interfaces. How is an amputee supposed to command the prosthesis what to do (i.e., how to grasp an object) and with what force (i.e., holding a hammer or grasping an egg)? So far, in literature, the most interesting results have been achieved by applying machine learning to forearm surface electromyography (EMG) to classify finger movements; but this approach lacks, in general, the possibility of quantitatively determining the force applied during the grasping act. In this paper we address the issue by applying machine learning to the problem of regression from the EMG signal to the force a human subject is applying to a force sensor. A detailed comparative analysis among three different machine learning approaches (Neural Networks, Support Vector Machines and Locally Weighted Projection Regression) reveals that the type of grasp can be reconstructed with an average accuracy of 90%, and the applied force can be predicted with an average error of 10%, corresponding to about 5N over a range of 50N. None of the tested approaches clearly outperforms the others, which seems to indicate that machine learning as a whole is a viable approach.

Accidents occurring with classical industrial robots often lead to fatal injuries. Presumably, this is to a great extent caused by the possibility of clamping the human in the confined workspace of the robot. Before generally allowing physical cooperation of humans and robots in future applications it is therefore absolutely crucial to analyze this extremely dangerous situation. In this paper we will investigate many aspects relevant to this sort of injury mechanisms and discuss the importance to domestic environments or production assistants. Since clamped impacts are intrinsically more dangerous than free ones it is fundamental to discuss and evaluate metrics to ensure safe interaction if clamping is possible. We compare various robots with respect to their injury potential leading to a main safety requirement of robot design: Reduce the intrinsic injury potential of a robot by reducing its weight.

The desired coexistence of robotic systems and humans in the same physical domain, by sharing their workspace and actually cooperating in a physical manner, poses the very fundamental problem of ensuring safety to the user. In this paper we will show the influence of robot mass and velocity during blunt unconstrained impacts with humans. Several robots with weights ranging from 15-2500 kg are impacted at different velocities with a mechanical human head mockup. This is used to measure the so-called head injury criterion, mainly a measure for brain injury. Apart from injuries indicated by this criterion and a detailed analysis of chest impacts we point out that e.g. fractures of facial bones can occur during collisions at typical robot velocities. Therefore, this injury mechanism which is more probable in robotics is evaluated in detail.

Currently, large efforts are undertaken to bring robotic applications to domestic environments. Especially physical human-robot cooperation is a major concern and various design and control methodologies were developed on the way to achieve this task. In particular, this necessitates the evaluation of injury risks a human is exposed to in case he is hit by a robot. In this video several blunt impact tests are shown, leading to an assessment of which factors dominate injury severity. We will illustrate the effect robot speed, robot mass, and constraints in the environment have on safety in human- robot impacts. It will be shown that the intuition of high impact loads being transmitted by heavy robots is wrong. Furthermore, the conclusion is induced that free impacts are by far less dangerous than being crushed.

Compliance in robot mounted force/torque sensors is useful for soft mating of parts. However it generates nearly undamped oscillations when moving the end-effector in free space. In this paper, input shaping control is investigated to damp such unwanted flexible modes. We present a new design technique that creates long impulse sequences to adapt input shaping to systems with long sampling period and to compensate the resulting time delay. This makes the method feasible for industrial robots. In addition to the conventional input shaping which causes oscillations to stop only after applying the last impulse, we also minimize the quadratic control error until this time step is reached.

A new method for the computation of the dynamics of structure-varying kinematic chains is proposed. This is based on the complete redefinition of the system connectivity deriving from a given structural change. The derived computational efficiency is then described with examples of typical motion planning tasks and structure changes for free-flying robots. These include open branched chains and closed loops, in free and grappled conditions. The method may then contribute to the efficiency of motion planning for robots which may require different kinematic structures for a given task.

This paper presents two novel approaches for accurate intrinsic and extrinsic camera calibration. The rationale behind them is the widespread violation of the traditional assumption that the metric structure of the calibration object is perfectly known. A novel formulation parameterizes a checkerboard calibration pattern in such a way that the calibration performs optimally irrespective of its actual dimensions. Simulations and experiments show that it is very rare for traditional calibration methods to come by the accuracy readily attained by this approach.

The object level control of a dexterous robot hand provides an intuitive high-level interface to solve fine manipulation tasks. In the past, many algorithms were proposed based on a weighted pseudoinverse of the grasp map. In a different approach Stramigioli introduces a virtual object - called "intrinsically passive controller (IPC)". This controller distributes the generalized object forces using coupling springs whose weighting have an intuitive physical meaning. Even though this controller has been known for several years we will present the first experimental results for a four-fingered hand. Furthermore, the term virtual grasp map is introduced and a method to parameterize the stiffness parameters in order to obtain an effective object level stiffness and a damping design is proposed. An implementation of the IPC is tested on the DLR Hand II and its performance is analyzed by manipulating soft and stiff objects.

Facing new tasks, the conventional rigid design of robotic joints has come to its limits. Operating in unknown environments current robots are prone to failure when hitting unforeseen rigid obstacles. Moreover, safety constraints are a major aspect for robots interacting with humans. In order to operate safely, existing robotic systems in this field are slow and have a lack of performance. To circumvent these limitations, a new robot joint with a variable stiffness approach (VS-Joint) is presented. It combines a compact and highly integrated design with high performance actuation. The VS- Joint features a highly dynamic stiffness adjustment along with a mechanically programmable system behavior. This allows an easy adaption to a big variety of tasks. A benefit of the joint is its intrinsic robustness against impacts and hard contacts, which permits faster trajectories and handling. Thus, it provides excellent attributes for the use in shoulder and elbow joints of an anthropomorphic robot arm.

This paper presents efficient computational algorithms of operational space dynamics for free-flying and for free-floating space robots. Due to the lack of the fixed base, the operational space formulation of the space robot is more complex than the fixed base robot system. By paying attention to this unique characteristic, however, the novel algorithm of the operational space dynamics for a single-serial-arm space robot is derived. Furthermore, by using the concept of the articulated-body system, recursive computation algorithms of the operational space formulation for a branching-arms on the space robot is developed. The realistic dynamic simulations are illustrated to verify the computational efficiency.

Walking is a fascinating way of locomotion that is very robust, especially in unstructured terrain. Many researchers devote their time to understanding its underlying principles and to build robots based on their findings. Using the fingers of DLR-Hand II a six-legged actively compliant walking robot is developed. It is intended to be used as testbed for the evaluation of different force- and position-based leg and gait control algorithms for hexapod walking in rough terrain. Following a brief overview of the finger hardware, the use of fingers as legs is analyzed and discussed. The body geometry as well as the systems constituting the robot are described. The compliance control algorithm used is explained and finally some experimental results are presented.

In the framework of physical Human-Robot Interaction (pHRI), methodologies and experimental tests are presented for the problem of detecting and reacting to collisions between a robot manipulator and a human being. Using a lightweight robot that was especially designed for interactive and cooperative tasks, we show how reactive control strategies can significantly contribute to ensuring safety to the human during physical interaction. Several collision tests were carried out, illustrating the feasibility and effectiveness of the proposed approach. While a subjective “safety” feeling is experienced by users when being able to naturally stop the robot in autonomous motion, a quantitative analysis of different reaction strategies was lacking. In order to compare these strategies on an objective basis, a mechanical verification platform has been built. The proposed collision detection and reactions methods prove to work very reliably and are effective in reducing contact forces far below any level which is dangerous to humans. Evaluations of impacts between robot and human arm or chest up to a maximum robot velocity of 2.7 m/s are presented.

Recent analysis on the stability boundary for haptic rendering assumed a stabilizing effect through a human operator holding a haptic device, without considering his/her dynamics directly. This paper derives stability boundaries of a linear model of a haptic system including those dynamics. It shows that all three elements of the human arm modeled as mass-spring-damper system contribute to stability. The haptic system itself is composed of a haptic device colliding with a virtual wall modeled as time-delayed discrete-time spring-damper system. Furthermore, the article proves that the recently found linear stability condition for haptic devices of Gil et al. still holds if a human is holding the haptic device. Finally, a relation to Colgatepsilas passivity condition defining a robustly stable region is given.

This paper presents a new developed multisensory five-fingered dexterous robot hand: the DLR/HIT Hand II. The hand has an independent palm and five identical modular fingers, each finger has three DOFs and four joints. All the actuators and electronics are integrated in the finger body and the palm. By using powerful super flat brushless DC motors, tiny harmonic drivers and BGA form DSPs and FPGAs, the whole fingerpsilas size is about one third smaller than the former finger in the DLR/HIT Hand I. By using the steel coupling mechanism, the phalanx distalpsilas transmission ratio is exact 1:1 in the whole movement range. At the same time, the multisensory dexterous hand integrates position, force/torque and temperature sensors. The hierarchical hardware structure of the hand consists of the finger DSPs, the finger FPGAs, the palm FPGA and the PCI based DSP/FPGA board. The hand can communicate with external with PPSeCo, CAN and Internet. Instead of extra cover, the packing mechanism of the hand is implemented directly in the finger body and palm to make the hand smaller and more human like. The whole weight of the hand is about 1.5Kg and the fingertip force can reach 10N.

In this paper we introduce a friction observer for robots with joint torque sensing (in particular for the DLR medical robot) in order to increase the positioning accuracy and the performance of torque control. The observer output corresponds to the low-pass filtered friction torque. It is used for friction compensation in conjunction with a MIMO controller designed for flexible joint arms. A passivity analysis is done for this friction compensation, allowing a Lyapunov based convergence analysis in the context of the nonlinear robot dynamics. For the complete controlled system, global asymptotic stability can be shown. Experimental results validate the practical efficiency of the approach.

The current discussion on physical human robot interaction and the related safety aspects, but also the interest of neuro-scientists to validate their hypotheses on human motor skills with bio-mimetic robots, led to a recent revival of tendon-driven robots. In this paper, the modeling of tendon-driven elastic systems with nonlinear couplings is recapitulated. A control law is developed that takes the desired joint position and stiffness as input. Therefore, desired motor positions are determined that are commanded to an impedance controller. We give a physical interpretation of the controller. More importantly, a static decoupling of the joint motion and the stiffness variation is given. The combination of active (controller) and passive (mechanical) stiffness is investigated. The controller stiffness is designed according to the desired overall stiffness. A damping design of the impedance controller is included in these considerations. The controller performance is evaluated in simulation.

For mobile manipulators envisioned in home environments a kitchen scenario provides a challenging testbed for numerous skills. Diverse manipulation actions are required, e.g. simple pick and place for moving objects and constrained motions for opening doors and drawers. The robot kinematics and link limits however are restrictive. Therefore especially a constrained trajectory will not be executable from arbitrary placements of the mobile manipulator. A two stage approach is presented to position a mobile manipulator to execute constrained linear trajectories as needed for opening drawers. In a first stage, a representation of a robot armpsilas reachable workspace is computed. Pattern recognition techniques are used to find regions in the workspace representation where these trajectories are possible. A set of trajectories results. In the second stage mobile manipulator placements are computed and the corresponding trajectories are checked for collisions. Compared to a brute force search through the workspace, the success rate of finding a mobile manipulator placement can be increased from 2% to 70%.

This paper presents a new developed multisensory five-fingered dexterous robot hand : the DLR/HIT Hand II. The hand has an independent palm and five identical modular fingers, each finger has three DOFs and four joints. All the actuators and electronics are integrated in the finger body and the palm. By using powerful super flat brushless DC motors, tiny harmonic drives and BGA form DSPs and FPGAs, the whole fingerpsilas size is about one third smaller than the former finger in the DLR/HIT Hand I. By using the steel coupling mechanism, the phalanx distalpsilas transmission ratio is exact 1:1 in the whole movement range. At the same time, the multisensory dexterous hand integrates position, force/torque and temperature sensors. The hierarchical hardware structure of the hand consists of the finger DSPs, the finger FPGAs, the palm FPGA and the PCI based DSP/FPGA board. The hand can communicate with external with PPSeCo , CAN and Internet. Instead of extra cover, the packing mechanism of the hand is implemented directly in the finger body and palm to make the hand smaller and more human like. The whole weight of the hand is about 1.5 Kg and the fingertip force can reach 10N.

In this paper, a novel type of impedance controllers for flexible joint robots is proposed. As a target impedance, a desired stiffness and damping are considered without inertia shaping. For this problem, two controllers of different complexity are proposed. Both have a cascaded structure with an inner torque feedback loop and an outer impedance controller. For the torque feedback, a physical interpretation as a scaling of the motor inertia is given, which allows to incorporate the torque feedback into a passivity-based analysis. The outer impedance control law is then designed differently for the two controllers. In the first approach, the stiffness and damping terms and the gravity compensation term are designed separately. This outer control loop uses only the motor position and velocity, but no noncollocated feedback of the joint torques or link side positions. In combination with the physical interpretation of torque feedback, this allows us to give a proof of the asymptotic stability of the closed-loop system based on the passivity properties of the system. The second control law is a refinement of this approach, in which the gravity compensation and the stiffness implementation are designed in a combined way. Thereby, a desired static stiffness relationship is obtained exactly. Additionally, some extensions of the controller to viscoelastic joints and to Cartesian impedance control are given. Finally, some experiments with the German Aerospace Center (DLR) lightweight robots verify the developed controllers and show the efficiency of the proposed control approach.

This video presents a humanoid two-arm system developed as a research platform for studying dexterous two-handed manipulation. The system is based on the modular DLR-Lightweight-Robot-III and the DLR-Hand-II. Two arms and hands are combined with a three degrees-of-freedom movable torso and a visual system to form a complete humanoid upper body. The diversity of the system is demonstrated by showing the mechanical design, several control concepts, the application of rapid prototyping and hardware-in-the-loop (HIL) development as well as two-handed manipulation experiments and the integration of path planning capabilities.

The influence of viscous damping and delay on the stability of haptic systems is studied in this paper. The stability boundaries have been found by means of different approaches. Although the shape of these stability boundaries is quite complex, a new linear condition which summarizes the relation between virtual stiffness, viscous damping and delay is proposed. This condition is independent of the mass of the haptic device. The theoretical results are supported by simulations and experimental data using the DLR light-weight robot.

We describe an efficient, reliable, and robust four-rotor flying platform for indoor and outdoor navigation. Currently, similar platforms are controlled at low frequencies due to hardware and software limitations. This causes uncertainty in position control and unstable behavior during fast maneuvers. Our flying platform offers a 1 kHz control frequency and motor update rate, in combination with powerful brushless DC motors in a light-weight package. Following a minimalistic design approach this system is based on a small number of low-cost components. Its robust performance is achieved by using simple but reliable highly optimized algorithms. The robot is small, light, and can carry payloads of up to 350g

In this paper, the dynamic model of a robot with antagonistic actuated joints is presented, and the problem of full linearization via static state feedback is analyzed. The use of transmission elements with nonlinear relation between the displacement and the actuated force allows to control both the position and the stiffness of each joint. The main advantage of this actuation modality is that the achieved stiffness becomes a mechanical characteristic of the system and it is not the result of an immediate control action as in the classical impedance control scheme (Davison, 2003). Different examples of implementation of this kind of devices are known in literature, even if limited to one single joint (Kjita et al., 2003; Laumond and Kineocam, 2006; Mansard and Chaumette, 2004 and 2006) and the application of antagonistic actuated kinematic chains in the field of robotic hand design is under investigation (Stasse et al., 2006). After a brief review of the dependence of the properties of antagonistic actuation on the transmission elements characteristics, a scheme for simultaneous stiffness-position control of the linearized system is presented. Finally, simulation results of a two-link antagonistic actuated arm are reported and discussed.

The paper addresses the automatic assembly of planar parts with complex geometry. Its main focus is on the automatic generation and parameterization of the assembly sequence, which should provide maximal robustness with respect to positioning errors of the robot and residual position uncertainties of vision based object localization. The assembly utilizes active or passive compliance of the robot in order to align the parts automatically. Success of the automatic alignment, i.e. the convergence of the assembly process can be guaranteed using the means of regions of attraction (ROA). The planning optimizes the assembly trajectories and parameters in such a way that the ROA is maximized for a given part geometry. For the convergence analysis, passivity properties of the robot and the environment are used. The method is validated through extensive experiments and can be successfully applied also for the automated assembly planning with passive compliance devices, as widely used today in industrial automation.

This paper deals with the concept and implementation of a multi-purpose vision platform. In robotics, numerous applications require perception. A multi-purpose vision platform suited for object recognition, cultural heritage preservation and visual servoing at the same time is missing. In this work, we draw attention to the design principles for such a vision platform. We present its implementation, the 3D-modeller. In specifying and combining multiple sensors, laser-range scanner, laser-stripe profiler and stereo vision, we derive the required mechanical and electrical hardware design. The concepts for synchronization and communication round offs our approach. Precision and frame rate are presented. We illustrate the versatility of the 3D-modeller by addressing four applications: 3D-modeling, exploration, tracking and object recognition. Due to its low weight and generic mechanical interface, it can be mounted on industrial robots, humanoids, or free-handed as well. The 3D-modeller is flexibly applicable, not only in research but also in industry, especially in small batch assembly.

The paper describes the modeling and control of a robot with flexible joints (the DLR medical robot), which has strong mechanical couplings between pairs of joints realized with a differential gear-box. Because of this coupling, controllers developed before for the DLR light-weight robots cannot be directly applied. The previous control approach is extended in order to allow a multi-input-multi-output (MIMO) design for the strongly coupled joints. Asymptotic stability is shown for the MIMO controller. Finally, experimental results with the DLR medical robot are presented.

The control of humanoid manipulators is very challenging due to the large number of degrees of freedom and the resulting redundancy. Using joint-level control complex planning algorithms are needed to accomplish tasks. For intuitive operation and hence short development times of applications high-level control interfaces are needed. Furthermore, for many tasks it is desirable to define an impedance behavior in task space. In this paper a flexible control law is proposed which offers object-level impedances for two-handed manipulation. The controller structure is based on the well-known compliance control law. The main contributions of this work are the way how to combine several potential functions for two-handed manipulation and the experimental validation of hand-arm coordination. The controller is implemented on DLR's humanoid manipulator Justin and its performance is demonstrated experimentally by unscrewing a can and motion of a grasped box.

For the 3D reconstruction inside historic buildings, we need a marker-free automatic registration approach to align different views together, because GPS does not work indoors and markers are not allowed to paste on the walls. This paper presents an automatic matching process, which employs a novel algorithm, Dynamic Matching Tree technique, for a fast and stable coarse-matching to achieve the automatic pre-alignment of two point clouds and uses modified ICP to do a fine matching efficiently. The whole process can be divided in the following stages: preprocessing, 2-View matching and N-View matching. To validate our method, various experiments has been done on reconstruction of historic sites and industrial objects.

With the rapid development of information and communication technology and widely spreading of internet, digital landscape becomes a high topic recently. This paper presents a landscape modeling- and visualizationssystem. The first part of this paper focuses on the real time 3D-model- lingsprocess out of large point-clouds. Diverse experiments have been done on the reconstruction of various famous regions of Bavaria with tourist features. And the second part concentrates on the interactive online visualizationssystem for massive meshes. To enable an efficient interactive online visualization of these large meshes, we convert them firstly with a multi-resolution hierarchy, and then display them progressively. As a preprocessing for the inter-active rendering, we generate a hierarchical tree of bounding spheres from triangle meshes and write it to disk. It will be used for view frustum culling, backface culling and level-of-detail control. We use a recursive algorithm for display. And the traverse-depth in the tree is decided by the projected size of the current node on the screen. By an interactive rendering, once the user stops moving the mouse, the scene will be redrawn with successively smaller thresholds until a size of one pixel is reached.

This paper addresses an on-line parameter adaptation for a momentum accumulation control of a space robot in the post-grasping of a tumbling target whose dynamic parameters are unknown a priori. The model inaccuracies in the target lead to an unexpected tumbling motion after grasping a target. It is desired to transfer the entire angular momentum to the reaction wheels as quickly as possible while stabilization trajectory of the robot-arm is tracked to avoid self-collision in the chaser-robot. Firstly, we derive a momentum control method from the angular momentum equation to accumulate the entire angular momentum into the reaction wheels. The parameter inaccuracies degrade the control performance. Then, an online adaptation law by using coupling force and momentum is proposed. A numerical simulation is carried out to verify the operational performance of the proposed method in the presence of model uncertainty.

This paper addresses an adaptive control for free-floating space robots in the presence of model uncertainty. Firstly, the operational space dynamics for a free-floating robot is derived with a novel, computationally efficient formulation. Then, by using the new formulation, we propose an adaptive control for a free-floating space robot to compensate the model uncertainty. For performance improvement, a composite adaptive control by combination of the trajectory error and the reaction force is further discussed. To verify the effectiveness of the proposed methods, a three-dimensional realistic numerical simulation is carried out.

This paper presents a new control strategy based on the time domain passivity control approach which copes with the active nature of delayed communication channels. Describing the system by means of network elements, the energy of the communication channel can be computed in real time and subsequently dissipated, thus providing stable operation. This is done bilaterally, since the system energy may flow from master to slave and from slave to master. The approach is accompanied with some experiments which validate the method.

The development of applications for multi-sensor data fusion typically faces heterogeneous hardware components, a variety of sensing principles and limited computational resources. We present a concept for synchronization and communication which tackles these challenges in multi-sensor systems in a unified manner. Here, a combination of hardware synchronization and deterministic software signals is promoted for global synchronization. Patterns of event-driven communication ensure that sensor data processing and evaluation are not bound to runtime constraints induced by data acquisition anymore. The combination of unified range and pose data description, event-driven communication, and global synchronization allows to build 3D sensing applications for various tasks. The proposed concept is implemented and evaluated for a variety of applications based on the DLR multisensory 3D-modeller. Extendability to other range and pose sensors is straightforward.

In this paper the effect of sensor quantization on the control performance of robotics actuators in the steadystate condition is considered. First, the existence of a limit cycle mode due to the limited sensor resolution in the systems with P-controller is shown in analogy to [1]. Because of the poor transient response of the P-controlled system the extension to the PD-controller is thereafter taken into consideration. A simple solution for limit cycles avoidance in terms of modification of controller structure is provided. The experimental data confirm the theoretical analysis for the robotics actuators.

Wheeled-mobile robots operating in human environments typically encounter small steps. Surmounting steps is normally not considered when determining peak torque needs, yet it can be the maximum requirement. This work looks at the statics and dynamics of this situation to determine the necessary peak torque. It finds that using a dynamic model that includes the wheel elasticity is essential for properly representing a real-world tire. When torque is increased using a step function, energy is stored in the tire - higher tire elasticity eases climbing. Knowledge of this phenomenon could facilitate the use of smaller actuators. The model is numerically integrated and results are found to agree with experiment.

Humans have at some point learned an abstraction of the capabilities of their arms. By just looking at the scene they can decide which places or objects they can easily reach and which are difficult to approach. Possessing a similar abstraction of a robot arm's capabilities in its workspace is important for grasp planners, path planners and task planners. In this paper, we show that robot arm capabilities manifest themselves as directional structures specific to workspace regions. We introduce a representation scheme that enables to visualize and inspect the directional structures. The directional structures are then captured in the form of a map, which we name the capability map. Using this capability map, a manipulator is able to deduce places that are easy to reach. Furthermore, a manipulator can either transport an object to a place where versatile manipulation is possible or a mobile manipulator or humanoid torso can position itself to enable optimal manipulation of an object.

The motion of a free-floating space robot is characterized by the principle of conservation of angular momentum. It is well known that these angular momentum equations are nonholonomic, i.e., are nonintegrable rate equations. If the base of the free-floating robot is partially actuated, it is difficult to attain trajectories of the joints that result in point-to-point motion of the entire robot system in its configuration space. However, if the drift-less system associated with the angular momentum conservation equations is shown to be differentially flat, point-to-point maneuvers of the free-floating robot in its configuration space can be constructed. However, an open research problem in the current literature is to show the property of differential flatness for a general space robot. The primary contributions of this paper are as follows: (i) study systematically the structure of the nonholonomic rate constraint equations of a free-floating open-chain space robot with arbitrarily oriented joints and two momentum wheels; (ii) establish the design conditions under which the system exhibits differential flatness; (iii) exploit these design conditions for point-to-point trajectory planning and control of the space robot

ROKVISS, the German new space robotics technology experiment, was successfully installed outside at the Russian Service Module of the International Space Station (ISS) during an extravehicular space walk at the end of January 2005. Since February 2005 a two joint manipulator can be operated from ground via a direct radio link. The aim of ROKVISS is the in flight verification of highly integrated modular robotic joints as well as the demonstration of different control modes, reaching from high system autonomy to force feedback teleoperation. A main goal of the experiment is the evaluation of the dynamical parameters (especially friction, motor constant and stiffness), as well as the monitoring of their evolution over the duration of the mission, in order to validate the long term performance of the system. The paper gives first a short overview of the experiment and in particular a description of the applied control structures. The main focus of the paper is on the joint parameter identification results obtained so far, during one year of operation

This paper presents a novel system for accurate placement of pedicle screws. The system consists of a new light-weight (<10 kg), kinematically redundant, and fully torque controlled robot. Additionally, the pose of the robot tool-center point is tracked by an optical navigation system, serving as an external reference source. Therefore, it is possible to measure and to compensate deviations between the intraoperative and the preoperatively planned pose. The robotic arm itself is impedance controlled. This allows for a new intuitive man-machine-interface as the joint units are equipped with torque sensors: the robot can be moved just by pulling/pushing its structure. The surgeon has full control of the robot at every step of the intervention. The hand-eye-coordination problems known from manual pedicle screw placement can be omitted

This paper addresses an impedance control for a free-floating space robot in the grasping of a tumbling target with model uncertainty. Firstly, the operational space dynamics for a free-floating robot is derived with a novel, computationally efficient formulation. Then, considering the grasped target as a disturbance force on the end-effector, the proposed control method is completely independent of the target inertial parameters and the end-effector can follow a given trajectory in the presence of model uncertainty. To verify the effectiveness of the proposed method, a three-dimensional realistic numerical simulation is carried out

This paper analyses the time domain passivity control approach in the time-delayed telepresence context, and proposes a method which provides stable operation. The passivity controller for the two-port network which is created by the bilateral control and communication elements in (J. H. Ryu, et al., May 2002) is shown to be not valid if a time delay is introduced in the communication channel. Classical stability analysis for the delayed system is presented and used as argument and benchmark for the proposed solution. Simulations and experimental results are discussed and compared with classical stability analysis

Mechatronic systems are reaching a new level of complexity, both for the single component and for overall systems making necessary a new software concept for the development and usage of such systems. Here we introduce the agile robot development (aRD) concept, which is a flexible, pragmatic and distributed software design to support and simplify the development of complex mechatronic and robotic systems. It gives easy access to scalable computing performance (even in hard realtime) and is motivated by the abstract view on a robotic system as being a decentral net of calculation blocks and communication links. We discuss design considerations and an implementation of this concept and demonstrate its performance with first applications

Based on under-actuated mechanism and coupling principle, a five-fingered, multi-sensory and biomechatronic prosthetic hand has been designed. The multi-DOF hand comprises 13 joints and is controlled by 3 motors. Actuated by only one motor, the thumb can move along a cone surface which is superior in the appearance. Also driven by one motor and transmitted by springs, the mid finger, the ring finger and the little finger can move simultaneously and envelop objects with complex shape. On the other hand, during the hand designation, the handsome appearance has been considered and its glove prototype has been designed. The hardware system and the sensory system have been developed. Through Bluetooth wireless protocol, the hand can be controlled by voice signal. Furthermore, it can also be controlled by electromyography (EMG) signal like most prosthetic hand in existence. It has been verified by experiments that the hand has strong capability of self-adaptation grasp and can accomplish precise and power grasp

Physical damping is increasing the z-width of haptic simulations. This paper derives the normalized stability boundaries for physically damped one degree of freedom haptic devices colliding with a virtual wall represented as spring-damper system. These boundaries are independent of the haptic device's mass and the sampling time. Furthermore, the dependency of the maximum stable virtual stiffness is discussed. Moreover, this paper illustrates that the passive region which is defined by Colgate's passivity condition is a subset inside the stable region for undelayed systems, but not for delayed systems

Diffuse and changing specifications for the design of light-weight robots result in high design costs for the desired robotic system, especially the electronic modules and related software drivers. To reduce those costs, we created a flexible robot platform, consisting of FPGA joint modules that are connected by a high speed communication. To fully exploit the hardware flexibility, we introduce a flexible signal-oriented hardware abstraction that is based on a signal flow oriented middleware (SFMiddleware). SFMiddleware enables the transparent integration of changing joint hardware functionality with robot control applications. Utilizing a static system specification approach, we benefit from the abstraction of a middleware without the typical overhead of common middleware implementations. Thus, we achieve a small run-time footprint and control cycles of more than 10 kHz

In this paper, a generic approach to optimize the design of an actuated carrier for the DLR multi-arm surgical system is presented. The carrier is attached to the ceiling of the operating room and provides additional degrees of freedom to the surgical robots with the purpose of automatic, optimal positioning of their bases as well as guaranteeing high stiffness. Standard workspaces of minimally invasive as well as open surgical procedures are considered and optimization criteria are derived. The minimum necessary degrees of freedom of the carrier are obtained as well as the optimal segment dimensions by use of an optimization with genetic algorithms

The dynamics and control of a cable-suspended, two-arm robotic system are developed for an entertainment application. One manipulator arm is controlled to fulfil a user defined task. The second arm is then controlled to compensate for the disturbances on the cable-suspended platform arising from the motion of the first. Model-based feedforward control, stemming from the momentum conservation equations of a free-floating robot, is developed for the motion compensation problem. Furthermore, due to model uncertainty, sensor-based feedback control is introduced, to account for undesired oscillatory motions of the system. The latter control problem reduces to the dissipation of the oscillatory energy of the system, by means of adequate robot control. Both control methods are implemented and tested on an experimental set-up

The increasing number of launched satellites per year, calls for solutions to keep free operational space for telecommunication systems in geo-synchronized orbit, as well as to avoid the endangering of space systems in LEO (low-Earth orbit) and of the public living in the habited parts on Earth. Examples for such dangerous stranded space systems in the past are Skylab and MIR. In the future, the uncontrolled and accidental de-orbiting of other huge satellites is expected, where parts of these will hit the surface of the Earth. A feasible way to handle such problems might be to enforce the operational requirement to use some dedicated residual fuel for a controlled de-orbiting, or in case of GEO (geostationary orbit), to lift the satellites at their end of life into the graveyard orbit. Despite these measures, malfunctions of solar generators, control systems or thrusters cannot be avoided. Therefore, on-orbit servicing (OOS) will be a mandatory and challenging topic for space robotics in the near future. The outcome of national German projects like ROTEX, ESS and GETEX/ETS-VII represent a know-how which can be directly applied for the development of OOS-robotic systems. Control structures and several possible operational modes are discussed within this paper. The recently started national project ROKVISS already provides the necessary space-qualified hardware as well as the very powerful telepresence operational mode. The paper will concentrate on a description of the ROKVISS mission

A robot manipulator sharing its workspace with humans should be able to quickly detect collisions and safely react for limiting injuries due to physical contacts. In the absence of external sensing, relative motions between robot and human are not predictable and unexpected collisions may occur at any location along the robot arm. Based on physical quantities such as total energy and generalized momentum of the robot manipulator, we present an efficient collision detection method that uses only proprioceptive robot sensors and provides also directional information for a safe robot reaction after collision. The approach is first developed for rigid robot arms and then extended to the case of robots with elastic joints, proposing different reaction strategies. Experimental results on collisions with the DLR-III lightweight manipulator are reported

In this paper the Cartesian compliance control of a manipulator mounted on a flexible base is considered. The proposed control law aims at achieving a desired stiffness and damping in Cartesian coordinates while taking account of the base flexibility. The controller does not use any measurement of the base motion, however a model of the base stiffness is required. For the closed loop system, asymptotic stability in case of free motion is proven. Furthermore, considering interaction tasks, it is shown that the controlled manipulator system has a useful passivity property

ROKVISS, Germany's newest space robotics technology experiment, was successfully installed outside at the Russian Service Module of the International Space Station (ISS) during an extravehicular space walk at the end of January 2005. Since February 2005 a two joint manipulator is operated from ground via a direct radio link. The aim of ROKVISS is the in flight verification of highly integrated modular robotic joints as well as the demonstration of different control modes, reaching from high system autonomy to force feedback teleoperation (telepresence mode). The experiment will be operated for at least one year in free space to evaluate and qualify intelligent light weight robotics components under realistic circumstances for maintenance and repair tasks as foreseen in upcoming manned and unmanned space applications in near future. This paper focuses in the telepresence control mode, its technology and first results from the space experiment ROKVISS

Classical position-based visual servoing approaches rely on the presence of distinctive features in the image such as corners and edges. In this contribution we exploit a hierarchical approach for object detection, initial-pose estimation, and real-time tracking based first on colour distribution and subsequently on the shape and texture information. The shape model of the object is not limited to surface primitives but allow for any free-form surface not subject to self-occlusion. We evaluate the approach as part of a handshake scenario where a 7-DoF robot takes a free moving object over from a human

This paper presents a calibration method for eye-in-hand systems in order to estimate the hand-eye and the robot-world transformations. The estimation takes place in terms of a parametrization of a stochastic model. In order to perform optimally, a metric on the group of the rigid transformations SE(3) and the corresponding error model are proposed for nonlinear optimization. This novel metric works well with both common formulations AX=XB and AX=ZB, and makes use of them in accordance with the nature of the problem. The metric also adapts itself to the system precision characteristics. The method is compared in performance to earlier approaches

Holding an object and manipulating it in 6D is a key application for multifingered robot hands. In the past many algorithms were proposed based on a weighted pseudoinverse of the grasp map combined with an internal force control. The majority of these algorithms require robust contact detection/tracking and switching controllers. Employing the virtual object introduced by Stramigioli we present an object-level control law. We define a novel virtual object frame based on the robot hand configuration. Our control law takes a desired object frame and desired grasping forces as input, it is passive, has an intuitive physical meaning, and stability is even given in case a finger looses contact with the object. A damping design as a function of the desired object stiffness and the combined hand-object inertia is presented. The performance of the controller is proven in two experiments implemented on the DLR Hand II

Autonomous service robots have to recognize and interpret their environment to be able to interact with it. This paper focuses on service tasks such as serving a glass of water where a humanoid two-arm-system has to acquire an object from the scene. A task planner should be able to autonomously discern the necessary actions to solve the task. In the process, a path planner can be used to compute motion sequences to execute these actions. To plan trajectories, the path planner requires a pair of configurations, the start and the goal configuration of the robot, to be provided e.g. by a task planner. This paper proposes a method to autonomously find the goal configurations necessary to acquire objects from the scene and thus makes an attempt to bridge the gap between task planning and path planning. The method determines where to grasp an object by analyzing the scene and the influence of obstacles on the intended grasp location. For the case where the goal object can not be grasped due to obstructing obstacles, a solution is proposed

A new five-fingered underactuated prosthetic hand control system is presented in this paper. The prosthetic hand control part is based on an EMG motion pattern classifier which combines VLR (variable learning rate) based neural network with wavelet transform and sample entropy. This motion pattern classifier can successfully identify flexion and extension of the thumb, the index finger and the middle finger, by measuring the surface EMG signals through three electrodes mounted on the flexor digitorum profundus, flexor pollicis longus and extensor digitorum. Furthermore, via continuously controlling single finger's motion, the prosthetic hand can achieve more prehensile postures such as power grasp, fingertip grasp, etc. The experimental results show that the classifier has a great potential application to the control of bionic man-machine systems because of its high recognition capability

Specular reflections pose a particular challenge to intensity-based tracking approaches. The proposed scheme substitutes specular areas with respect to local image structures, extracted by the structure tensor. It can be applied to images of the beating heart prior to tracking and is shown to improve the estimation of local motion by natural landmarks significantly. The runtime complexity of the substitution scheme is analysed and time measurements of the different components show that it is well suited for realtime tracking.

The paper presents an impedance controller for a class of underactuated Euler-Lagrange systems based on energy shaping. For shaping the potential energy, feedback variables are introduced, which are functions of the collocated state variables only, but which are statically equivalent to the noncollocated state variables. In this way the passivity of the system can be ensured, while exactly satisfying all steady state requirements formulated in terms of the noncollocated states (such as desired stiffness and desired equilibrium configuration). The method is constructive, allowing the direct formulation of the controller and of the corresponding energy function. The controller additionally uses the noncollocated feedback to shape the kinetic energy. Under some conditions on the potential energy of the plant, the closed loop system can then be seen as a feedback interconnection of passive systems and is proven to be asymptotically stable. Experimental results for a flexible joint robot validate the proposed controller.

When industrial robot arms are controlled using sensor data the performance is dependent on the sensor sampling rate, on delays in signal processing, and on the robot dynamics. The paper presents an approach in which control is inherently stable as long as the time instant of sensing is known, independently of delays. In addition to sensor data the method uses the actual robot pose to compute a desired pose which is then controlled by the existing positional control loop. Updated sensor data affect the system as a refined target for positional control. So the positional control and the use of sensor data are decoupled. This is useful for the integration of a priori information on the task. The method is applicable especially for force control tasks as contour following and for visual servoing.

For precise control of robots along paths which are sensed online it is of fundamental importance to have a calibrated system. In addition to the identification of the sensor parameters - in our case the camera calibration - we focus on the adaptation of parameters that characterize the integration of the sensor into the control system or the application. The most important of such parameters are identified best when evaluating an application task, after a short pre-calibration phase. The method is demonstrated in experiments in which a robot arm follows a curved line at high speed.

Minimally invasive surgery (MIS) challenges the surgeon’s skills due to his separation from the operation area which can only be reached with long instruments. To overcome these drawbacks, minimally invasive robotic surgery (MIRS) plays an important role. This paper describes the development of actuated and sensorized instruments for minimally invasive robotic surgery which help to increase the surgeon’s immersion and dexterity.

This paper presents a surface Electromyography (EMG) motion pattern classifier which combines Levenberg-Marquardt (LM) based neural network with parametric Autoregressive (AR) model. This motion pattern classifier can successfully identify three types of motion of thumb, index finger and middle finger, by measuring the surface EMG through two electrodes mounted on the flexor digitorum profundus and flexor pollicis longus. Furthermore, via continuously controlling single finger’s motion, the five-fingered underactuated prosthetic hand can achieve more prehensile postures such as power grasp, centralized grip, fingertip grasp, cylindrical grasp, etc. The experimental results show that the classifier has a great potential application to the control of bionic man-machine systems because of its fast learning speed, high recognition capability and strong robustness.

In this paper, FPGA (field programmable gate array) based hardware architecture for the HIT/DLR hand has been investigated. With the FPGAs for lower level control and DSP (digital signal processor) for higher level control, the whole hardware is very intelligent. By using the high capacity of FPGAs, the additional hardware such as communication controller and PWM generators, can be implemented in a single chip and the hardware system is more flexible and compact. In each finger there is an FPGA for data collection, brushless DC motors control and communication with palm's FPGA by point-to-point serial communication (PPSeCo). The kernel of the hardware system is a PCI-based high speed floating-point DSP for data processing, and FPGA for high-speed (up to 25Mbps) real-time serial communication with the palm's FPGA. There needs only 4 cables for the data transmission and the sampling cycle for each sensor is only 200 /spl mu/s. This paper presents the basic ideas behind the HIT/DLR hand's hard- and software architecture adapted to new needs in data processing.

The paper presents a Cartesian impedance controller for flexible joint robots based on the feedback of the complete state of the system, namely the motor position, the joint torque and their derivatives. The approach is applied to a quite general robot model, in which also a damping element is considered in parallel to the joint stiffness. Since passivity and asymptotic stability of the controller hold also for varying damping matrices, some possibilities of designing those gain matrices (depending on the actual inertia matrix) are addressed. The passivity of the controller relies on the usage of only motor side measurements for the position feedback. A method is introduced, which provides the exact desired link side stiffness based on this motor position information. Experimental results are validating the proposed controller.

For the evaluation of grasp quality, different measures have been proposed that are based on wrench spaces. Almost all of them have drawbacks that derive from the non-uniformity of the wrench space, composed of force and torque dimensions. Moreover, many of these approaches are computationally expensive. We address the problem of choosing a proper task wrench space to overcome the problems of the non-uniform wrench space and show how to integrate it in a well-known, high precision and extremely fast computable grasp quality measure.

The paper presents hardware and software architectures of the HIT/DLR Hand. The hand has four identical fingers and an extra degree of freedom (d.o.f) for the palm. In each finger, there is a re-configurable Field Programmable Gate Array (FPGA) for data acquisition, Brushless DC (BLDC) motor control and communication with the palm's FPGA by Point-to-Point Serial Communication (PPSeCo). The kernel of the hardware system is a PCI-based high speed floating-point Digital Signal Processor (DSP) for data processing, and an FPGA for high speed (up to 25 Mbps) real-time serial communication with the palm's FPGA. In order to achieve high modularity and reliability of the hand, a fully mechatronic integration and analog signals in-situ digitalization philosophy are implemented to minimize the dimension, number of the cables (5 cables including power supply) and protect data communication from outside disturbances. Furthermore, according to the hardware architecture of the hand, a hierarchical software architecture has been established to perform all data processing and control of the hand. The software structure provides basic Application Programming Interface (API) functions and skills to access all hardware resources for data acquisition, computation and teleoperation.

In this paper a novel approach to the Cartesian impedance control problem for robots with flexible joints is presented. The proposed controller structure is based on simple physical considerations, which are motivating the extension of classical position feedback by an additional feedback of the joint torques. The torque feedback action can be interpreted as a scaling of the apparent motor inertia. Furthermore the problem of gravity compensation is addressed. Finally, it is shown that the closed loop system can be seen as a feedback interconnection of passive systems. Based on this passivity property a proof of asymptotic stability is presented.

This paper presents the DLR Laser Stripe Profiler as a component of the DLR multisensory Hand-Guided Device for 3D modeling. After modeling the reconstruction process, we propose a novel method for laser plane self-calibration based on the assessment of the deformations the miscalibration leads to. In addition, the requirement for absence of optical filtering implies the development of a robust stripe segmentation algorithm. Experiments demonstrate the validity and applicability of the approaches.

The concept of C-space entropy as a measure of knowledge of C-space for sensor-based path planning and exploration for general robot-sensor systems was introduced in Yu, Y. and Gupta, K. (2000). The robot plans the next sensing action to maximally reduce the expected C-space entropy, also called the maximal expected entropy reduction, or MER criterion. The expected C-space entropy computation, however, made an idealized assumption. The sensor was assumed to measure exact data, i.e., it was not subject to noise. In this paper we extend this approach by using a real noisy sensor model. Sensing actions can then be compared on the basis of their uncertainty models. This offers the ability for using more than one principle sensor (multisensory exploration), because sensor readings can be weighted by evaluating the expected measurement quality. Additionally, it makes robot motion planning viable for tasks such as object surface inspection, which require the robot to come very close to the obstacles to achieve high sensing accuracy.

In the field of service robotics, whole arm contact with an unstructured environment or human beings becomes a major issue. Therefore soft robots, which mean robots with passively (or mechanically) compliant joints, become more and more important. In this work we analyze what Cartesian stiffness at the tool center point one can achieve with a passively compliant, redundant robot with variable joint stiffness. We restrict this work to the special case of uncoupled joint stiffness only, as coupling of joint stiffness seems to be mechanically difficult to realize. Finally we discuss a Cartesian controller, which incorporates the compliance of the joints and ensures the correct stiffness behavior also for high displacements from the desired position.

This work describes a data fusion architecture for robotic assembly tasks based on human sensory-motor skills. These skills are transferred to the robot through geometric and dynamic perception signals. Artificial neural networks are used in the learning process. The data fusion paradigm is addressed. It consists of two independent modules for optimal fusion and filtering. Kalman techniques linked to stochastic signal evolutions are used in the fusion algorithm. Compliant motion signals obtained from vision and pose sense are fused, enhancing the task performance. Simulations and peg-in-hole experiments are reported.

This paper addresses the problem of impedance control for flexible joint robots based on a singular perturbation approach. Some aspects of the impedance controller, which turned out to be of high practical relevance during applications are then addressed, such as the implementation of nullspace stiffness for redundant manipulators, the avoiding of mass matrix decoupling and the related design of the desired damping matrix. Finally, the proposed methods are validated through measurements on the DLR robot.

At our institute, two generations of antropomorphic hands have been designed. In quite a few experiments and demonstrations we could show the abilities of our hands and gain a lot of experience in what artificial hands can do, what abilities they need and where their limitations lie. In this paper, we would like to give an overview over the experiments performed with the DLR hands, our hands abilities and the things that need to be done in the near future.

This paper presents the current work progress of HIT/DLR Dexterous Hand. Based on the technology of DLR Hand II, HIT and DLR are jointly developing a smaller and easier manufactured robot hand. The prototype of one finger has been successfully built. The finger has three DOF and four joints, the last two joints are mechanically coupled by a rigid linkage. All the actuators are commercial brushless DC motors with integrated analog Hall sensors. DSP based control system is implemented in PCI bus architecture and the serial communication between the hand and DSP needs only 6 lines(4 lines power supply and 2 lines communication interface). The fingertip force can reach 10N.

In the robotic community more and more hands have been developed. These newly designed manipulators greatly outperform their ancestors in terms of available sensor signals, applicable grasping force, mechanical stability, reliability, kinematic design and more. This development extends the possible range and complexity of applications of robotic grippers also to areas outside of well structured laboratories and simple tasks. It also calls for more flexible control structures to provide a framework for implementing and executing these newly arising tasks without having to start from scratch for each new task. During the last few years we developed a control system architecture for DLR hand II that proved to be useful for a great variety of different applications. This paper presents the basic ideas behind DLR hand II's hard- and software architecture adapted to new needs in data processing.

Autonomously grasping a predefined object is a topic of recent research in the field of service robotics. On the one hand, there are numerous approaches in the area of image processing concerned with recognition and localization of this object. On the other hand, a lot of work has been done in the development of planning, approaching and grasping of the object with a dextrous manipulator mounted on top of a robot arm. However, in-between locating and grasping, there are significant sources of uncertainty, e.g. estimation errors in image processing, errors in calibration of cameras and robot alone and with respect to each other, and positioning errors in the robot control. During the critical closing phase of grasping however, visual servoing and position correction is almost impossible to achieve due to obstruction of the object by the gripper. This paper presents an algorithm to locally estimate the position and orientation of the object to be grasped from contact information and a geometric description of the object. In this scenario, an object description is usually available to a sufficiently accurate extent from grasp planning.

This paper addresses the problem of motion planning for free-flying robots. Full state actuation is considered to allow for large displacements of the spacecraft. Motion planning is formulated as an optimization problem and kinematic as well as dynamic constraints are considered. The chosen optimization criteria are spacecraft actuation and final time. The proposed method allows solutions which do not require any spacecraft actuation for those end goals for which the robot motion is sufficient.

This paper addresses the impedance control problem for flexible joint manipulators. An impedance controller structure is proposed, which is based on an exact decoupling of the torque dynamics from the link dynamics. A formal stability analysis of the proposed controller is presented for the general tracking case. Preliminary experimental results are given for a single flexible joint.

A force-feedback actuation loop for a steer-by-wire vehicle is developed. It is shown that the performance of this loop can be essentially improved by the introduction of a torque sensor. Model reference based control algorithms based on disturbance observer (DOB) and active observers (AOB) are applied to enhance the robustness vs. non-modelled dynamics and uncertain driver impedance.

In this work, a cartesian impedance controller purposely designed for dexterous manipulation is described. Based on the main features of the DLR Hand II, concerning kinematic structure and sensory equipment of fingers, this control strategy allows to overcome the main problems encountered in fine manipulation, namely: effects of the friction (and unmodeled dynamics) on robot performances and occurrence of singularity conditions. The achieved control scheme bas been experimentally validated by testing it on a finger of the DLR Hand.

Many methods for generating and analyzing grasps have been developed in the recent years. They gave insight and comprehension of grasping with robot hands but many of them are rather complicated to implement and of high computational complexity. In this paper we study if the basic quality criterion for grasps, the force-closure property, is in principle easy or difficult to reach. We show that it is not necessary to generate optimal grasps, due to a certain quality measure, for real robot grasping tasks where an average quality grasp is acceptable. We present statistical data that confirm our opinion that a randomized grasp generation algorithm is fast and suitable for the planning of robot grasping tasks.

In this paper a modification of the time domain passivity controller is presented to improve its performance and transparency in case of multi degrees of freedom (dof) haptic interaction. In multi-dof application the concept needs to be extended by additional conditions to distribute the adaptive damping appropriately among the degrees of freedom. This can be solved by using the geometrical information coded in the output signals of the system. Experiments show the validity of this concept.

A recently proposed method for stabilizing haptic interfaces and teleoperation systems was tested with a "PHANToM" commercial haptic device. The "passivity observer" (PO) and "passivity controller" (PC) stabilization method was formed to stabilize the system but also excite high frequency mode in the device. To solve this problem, we propose a method to use a time-varying desired energy threshold instead of fixed zero energy threshold for the PO, and make the actual energy input follow the time-varying energy threshold. With the time-varying energy threshold, we make the PC control action smooth without sudden impulsive behavior by distributing the dissipation. The proposed new PO/PC approach is applied to PHANToM with high stiffness (K=500 N/m), and stable and smooth contact is guarantee. Resetting and active environment display problems can also be solved with the reference energy following idea.

For a class of robot arms, a proportional-derivative (PD) controller plus gravity compensation yields the global asymptotic stability for regulation tasks, and some proportional-integral-derivative (PID) controllers guarantee local regulation without gravity cancellation. However, these controllers cannot render asymptotic stability for tracking tasks. In this paper, a simple decentralized continuous sliding PID controller for tracking tasks that yields semiglobal stability of all closed-loop signals with exponential convergence of tracking errors is proposed. A dynamic sliding mode without reaching phase is enforced, and terminal attractors, as well as saturated ones, are considered. A comparative experimental study versus PD control, PID control, and adaptive control for a rigid robot arm validates our design.

The paper compares various approaches to implementing a compliant Cartesian behavior for robotic manipulators: impedance, admittance and stiffness control. A new controller structure is proposed, which consists of an impedance controller enhanced by local stiffness control. This structure consistently takes into account the two time scale property of the joint and Cartesian control loops. The DLR light-weight robot, with its position, torque and impedance interfaces on joint level, is an adequate platform for the implementation of the presented methods. The experimental results are discussed and a critical comparison of the performance with different controllers is made. As an application for the new control structure the fast and intuitive teaching of an insertion task (piston into a motor block) is described.

Most of the algorithms used in grasp planning, force optimization and control of multifingered hands need information about the points of contact with a grasped object as well as the normal of the surface in the contact point. With no image processing, this information is gained from tactile sensor arrays, multidimensional force/torque sensors or a priori knowledge. This paper presents a method for those cases, when no good quality sensors are present or measurements are to be improved by sensor fusion. An algorithm is developed to determine the contact points and inherently the surface normal from only joint angle sensors and a geometric description of the fingertip in the 3D case. This is done by observing the constrained motion of fingers securely grasping an object.

A third generation of torque-controlled light weight robots has been developed in DLR's robotics and mechatronics lab which is based on all the experiences that have been had with the first two generations. It aims at reaching the limits of what seems achievable with present day technologies not only with respect to light-weight, but also with respect to minimal power consumption and losses. One of the main gaps we tried to close in version III was the development of a new, robot-dedicated high energy motor designed with the best available techniques of concurrent engineering, and the renewed efforts to save weight in the links by using ultralight carbon fibres.

A multimodal Internet-based telepresence system with interactive stereo vision, additional photo-realistic predictive display, and kinesthetic bilateral coupling in three degrees of freedom is presented. For easy coupling of various telepresence components by a unified interface structure and a generic link management protocol a CORBA based interconnection framework has been developed. By solving a peg-in-hole task in an industrial environment and with varying communication delay times, the system provides all capabilities for multimodal telepresent remote maintenance and assembly.

Usually, grasp planning can be split into two phases. In the first phase one tries to find a set of contacts that allow for stable grasping of an object. In the second phase a feasible hand pose that realizes the grasp with a given hand is calculated. While this point is important for a practical grasp planning system, it has either been considered trivial or been solved by crude heuristics in most cases. Here we present an approach for calculating the hand and finger pose for a given grasp. The problem is formulated as a constraint satisfaction problem and then solved using optimization techniques. The method is applied to two different grasp types: the well known precision grasp and the pinch grasp which is preferred by men when grasping small objects.

The paper discusses new developments of the data fusion paradigm due to Cortesao and Koeppe (1999, 2000). A bank of Kalman filters is analyzed in the fusion process. Experiments for a robotic compliant motion task (peg-in-hole) emerged from human skills are reported. Stereo vision and pose sense are fused to execute the task. Feedforward artificial neural networks (ANNs) are trained to transfer human skills to robotic manipulators.

This paper addresses the adaptive tracking control of flexible joint manipulators based on the singular perturbation approach. Comparisons between adaptive and nonadaptive tracking controllers are given. Experimental results with the DLR's light weight robot are given for the 2-DOF case.

We propose a method for identifying the parameters of a flexible joint robot based on the motor position, current and the additional joint torque sensor information. We make some theoretical remarks on the passivity property of the new controller used in the experiments. Simulation results for the movement of the complete robot are compared to measurements in order to validate the quality of the model and the performance of the controller.

This paper outlines the 2nd generation of multisensory hand design at DLR, based on the results of the DLR Hand I we analysed. An open skeleton structure for better maintenance with semi-shell housing and the new automatically reconfigurable palm have been equipped with more powerful actuators to reach 30 N on the fingertip. The newly designed sensors as the 6-DOF fingertip force torque sensor, the integrated electronics and the new communication architecture with a reduction of cabling to the hand to only 12 lines, are outlined. The Cartesian impedance control of all the fingers completes the new 13-DOF hand.

The paper describes the recent design and development efforts in DLR Robotics Lab towards the second generation of light-weight robots. The design of the light weight mechanics, integrated sensors and electronics is outlined. The fully sensory joint, with motor and link position sensors as well as joint torque sensors enables the implementation of effective vibration damping and advanced control strategies for compliant manipulation. The mechatronic approach incorporates a tight collaboration between mechanics, electronics and controller design. The authors hope that important steps towards a new generation of service and personal robots have been achieved.

The theory of active observers was initially described by Cortesao et al. (2000). This paper introduces properties of the Kalman gains used in the active observer (AOB) design. An important result of the state-space design is demonstrated, which allows stiffness adaptation without changing the control structure. Experiments with a human-robot skill transfer system to perform the peg-in-hole compliant motion task are described, showing the importance of the AOB.

We present a system for catching a flying ball with a robot arm using off-the-shelf components (PC based system) for visual tracking. The ball is observed by a large baseline stereo camera, comparing each image to a slowly adapting reference image. We track and predict the target position using an extended Kalman filter, also taking into account the air drag. The calibration is achieved by simply performing a few throws and observing their trajectories, as well as moving the robot to some predefined positions.

In the growing field of service robotics the interaction between humans and robots is an important topic. In this paper the interactive features of the kinematically redundant DLR lightweight robot axe presented. At the example of an inverted pendulum the "interactive nullspace motion" is introduced, where the user is able to modify the configuration as a subtask while balancing the pendulum as primary task. Different ways of nullspace interaction are shown, either contact-free by a teach-device or a position tracker, or by touching the robot, whereas the joint torque sensors measure the external touch.

Describes design and development efforts in DLR's robotics lab towards a new generation of ultra-light weight robots with articulated hands. The design of fully sensorized joints with complete state feedback and the underlying mechanisms are outlined. The second light-weight arm generation is available now, as well as the second generation of a worldwide most highly integrated 4 finger-hand is available now. Thus we hope that important steps towards a new generation of service and personal robots have been achieved.

Outlines the situation in orbital space robotics with special reference to what DLR (German Aerospace Center) has contributed to the field. After our ROTEX experiment, the first remotely controlled space robot inside the space shuttle, the Japanese ETS VII has now been the first remotely controlled free-flying space robot. We had the opportunity to control this arm from the ground, too, including the use of the robot arm as a satellite attitude controller. It is outlined how it is now time to take the next steps towards operational ground-controlled space robot systems, presumably first on the International Space Station, but later on as free flying robonauts assisting or even replacing extra vehicular activities.

Recent research in industrial robotics aims at the involvement of additional sensory devices to improve robustness, flexibility and performance of common robot applications. Many different sensors have been developed over the past years to fit the requirements of different but very specific tasks. Special seam tracking sensors support the robot in welding applications. Vision systems are common in quality control and inspection. Force/torque sensors mounted to a robot's wrist are still an exception and limited to the fields of scientific research. Compared to the number of annual robot sales the number of sensor equipped robots is still negligible, although the benefits of sensory feedback are obvious. In this paper we introduce a general approach to tackle the problem of sensor-based robot assembly. We realized a flexible assembly cell that includes a variety of different sensors for mating with moving parts.

Addresses the problem of controlling the joints of a flexible joint robot with a state feedback controller and proposes a gradual way of extending such a controller towards feedback linearization. The global asymptotic stability for the state feedback controller with gravity compensation is proven. Experimental results on the DLR light-weight robots validate the method.

The article describes a systematic procedure to design a force controller with active observers (AOB). The design is based on pole placement using discrete state space theory. The concept of AOB is introduced as a starting point to perform robust estimates of the system state. The robustness of the system is accomplished through optimal noise processing embedded in the control strategy. The design was tested as a force controller in a Manutec R2 Robot at the DLR.

Architectures for the execution of autonomous compliant motion (ACM) require modules for 6 degree-of-freedom localization, navigation and force control of robot manipulators. We review existing model- and signal-based approaches to ACM and present a solution for the synthesis of localization and navigation using sensorimotor signals recorded from human demonstration.

Telepresence in minimally invasive surgery (MIS) is a promising application for robotics because the robot enhances the manipulation and sensation capabilities of the surgeon. Many of these surgical robots are equipped with passive joints that guarantee that no forces are exerted to the entry point. Due to these passive joints, which are dependent on the entry point's position, new algorithms for cartesian control have to be developed. After introducing the kinematics of the robot, used in the DLR minimally invasive robot surgery scenario, we show how the entry point (trocar) can be estimated, followed by the calculation of the inverse kinematics. To allow cartesian velocity control of the robot system we build an observer that is used to close the velocity control loop. Experiments show the performance of the chosen algorithms.

In our previous work, several system elements for space teleoperation were introduced: a Singularity Consistent (SC) teleoperation system, a mixed force and motion commands based space teleoperation system, and a 6-DOF haptic interface (Y. Tsumaki and M. Uchiyama, 1997; Y. Tsumaki et al., 1998). We introduce an advanced space teleoperation system which includes all the above elements. As a result, the system realizes safety against the singularities and excessive force, with good perception for the operator. In addition, to verify the effectiveness of this system, we conducted experiments through the Internet between DLR in Germany and Tohoku University in Japan in January 1999.

We present a method for object recognition and pose estimation from noisy range data as provided by stereo processing. From the range dates, points of high surface curvature are estimated. By comparing three-point geometric relations, hypothetical correspondences are established between data and model points of high curvature. The hypothetical correspondences give rise to pose hypotheses which are evaluated with respect to the raw range data using a crude surface model. We show examples that demonstrate the method's tolerance to noise and occlusion.

The authors consider the cerebellum in the vertebrate motor control system. Analyzing the delays in this control loop as well as the complexity of the dynamics of the skeletomuscular system, we find that two well-known interpretations of the function of the cerebellum (that of a Smith predictor, or of an inverse model controller) are insufficient to solve the control problem at hand.

In order to grasp and manipulate real world objects, grasp planning systems are required, and they have to be very fast in order to be integrated in online planning systems for robots. This paper presents a method to compute a desirable grasp quality measure very fast and accurately. Based on this measure a heuristic approach towards fast planning of precision grasps for arbitrarily shaped 3D objects is described. A number of feasible grasp candidates are generated heuristically. These grasp candidates are qualified using the described grasp quality measure and the best candidate is chosen. The planned grasps are robust with respect to grasp placement. It is shown that only a relatively small number of grasp candidates has to be generated in order to obtain a good, although not optimal, grasp.

An adaptive architecture for feedforward control of industrial robots with standard positional controller is presented. This approach explicitly considers robots with elastic joints. It assumes that the real position of the tool centre point (TCP) can be recorded for offline evaluation. Compensation of elasticity is trained in a feedforward controller. Adaptation takes place without any knowledge of the physical system model. The performance of the method is demonstrated in experiments with a 6-axis industrial robot KUKA KR6/1 for which real path errors during full speed motion are reduced by 70%. Reductions of 50% can be expected for untrained paths or other robots of the same type.

Most of the approaches proposed in the field of sensor based assembly use force and velocity, i.e., sensor signals representing dynamics, as a base on which a desired compliant motion strategy is specified. In contrast, especially while performing alignment and insertion tasks the problem of sensing the assembly state of two objects w.r.t. each other is of geometric nature. Using a generic sensorimotor skill model for compliant motion tasks we discuss the difference between dynamic and geometric perception. We propose approaches of kinesthetic and visual sensing and show the robustness of compliant motion skill maps obtained from sensor signals representing the task geometry rather than the task dynamics.

Presents a novel Cartesian impedance control for the DLR (German Aerospace Center) Hand based on joint torque measurements. The fingertip appears as mechanical impedance when it contacts with an unknown obstacle. The impedance parameters can be adjusted in a certain range as needed in any Cartesian coordinate system. There is no switching mode between the motions in the free space and in the constraint environment. The paper also gives a detailed analysis of the finger's kinematics and dynamics model. Experimental results have verified the effectiveness and robustness of the proposed scheme.

A solution to the singularity problem of a non-redundant robot is proposed by reformulating the inverse kinematic problem as a constraint optimization problem. The main idea is to allow a cartesian error in a given subspace in the vicinity of a singularity and to minimize this error subject to operational constraints such as maximum motor speeds. As a result, in every sampling instant a series of linear least squares problems with linear equality and inequality constraints have to be solved. This task can be carried out on a Pentium processor within a few milliseconds. The new method is demonstrated by real experiments with an industrial robot.

Conventional computer vision methods for determining a robot's end-effector motion based on sensory data needs sensor calibration (e.g., camera calibration) and sensor-to-hand calibration (e.g., hand-eye calibration). This involves many computations and even some difficulties, especially when different kinds of sensors are involved. In this correspondence, we present a neural network approach to the motion determination problem without any calibration. Two kinds of sensory data, namely, camera images and laser range data, are used as the input to a multilayer feedforward network to associate the direct transformation from the sensory data to the required motions. This provides a practical sensor fusion method. Using a recursive motion strategy and in terms of a network correction, we relax the requirement for the exactness of the learned transformation. Another important feature of our work is that the goal position can be changed without having to do network retraining. Experimental results show the effectiveness of our method.

In this paper, we propose a new solution to the stereo correspondence problem by including features an intensity based matching. The features we use are intensity gradients in both the x and y directions of the left and the deformed right images. Although a uniform smoothness constraint is still used, it is nevertheless applied only to non-feature regions. To avoid local minima in function minimization, we propose to parameterize the disparity function by hierarchical Gaussians. A simple stochastic gradient method is used to estimate the Gaussian weights. Experiments with various real stereo images show robust performances.

The main features of DLR's dextrous robot hand as a modular component of a complete robotics system are outlined in this paper. The application of robotics systems in unstructured servicing environments requires dextrous manipulation abilities and facilities to perform complex remote operations in a very flexible way. Therefore we have developed a multisensory articulated four finger hand, where all actuators are integrated in the hand's palm or the fingers directly. It is an integrated part of a complex light-weight manipulation system aiming at the development of robonauts for space. After a brief description of the hand and it's sensorial equipment the hard- and software architecture is outlined with particular emphasis on flexibility and performance issues. The hand is typically controlled through a data glove for telemanipulation and skill-transfer purposes. Autonomous grasping and manipulation capabilities are currently under development.

We present a setup to control a four-finger anthropomorphic robot hand using a dataglove. To be able to accurately use the dataglove we implemented a nonlinear learning calibration using a novel neural network technique. Experiments show that a resulting positioning error not exceeding 1.8 mm, but typically 0.5 mm, per finger can be obtained; this accuracy is sufficiently precise for grasping tasks. Based on the dataglove calibration we present a solution for the mapping of human and artificial hand workspaces that enables an operator to intuitively and easily telemanipulate objects with the artificial hand.

A predictive architecture is presented to react on sensor data in the case of high speed motion and low bandwidth sensor data. This concept is used for the vision based control of an industrial robot to track a contour at a speed of 1.6 m/s. The vision task can be performed very fast since only 2 rows of the image are analyzed. In this way an accuracy of 0.3 mm is reached in spite of uncertainties in robot's kinematic parameters. Vision and control work asynchronously so that even delay times are tolerable during sensing as long as the time-instant of the exposure is known.

Gives a brief description of feedback control systems engaged in DLR's recently developed multisensory 4-finger robot hand. The work is concentrated on constructing the dynamic model and the control strategy for one joint of the fingers. One goal is to make the hand follow a dataglove for fine manipulation tasks. Our proposed strategy for this task is parallel torque/position control; sliding mode control is realized for the robust trajectory tracking in free space; while impedance control is provided for compliance control in the constrained environment; and an easily-designed parallel observer is used for the switch between these two control modes during the transition from or to contact motion. Some experimental results show the effectiveness of proposed strategy for the pure position control, torque control, and the transition control.

Compared with industrial robots, the DLR lightweight robot characterizes a new milestone in the field of robot design. The module structure, the highly integrated electrical and mechanical components, the very low own-weight/load ratio, the high motion speed as well as the joint-torque feedback capability are the main features of this new robot generation aiming towards the functionality of a human arm. The paper attempts to integrate the newest results in the field of robotics, nonlinear control theory and electric drive systems to formulate the possible solutions of the dynamic control issues, Some experimental or simulation results are given to confirm the effectiveness of the proposed control approaches.

We propose a stereo correspondence method by minimizing intensity and gradient errors simultaneously. In contrast to conventional use of image gradients, the gradients are applied in the deformed image space. Although a uniform smoothness constraint is imposed, it is applied only to nonfeature regions. To avoid local minima in the function minimization, we propose to parameterize the disparity function by hierarchical Gaussians. Both the uniqueness and the ordering constraints can be easily imposed in our minimization framework. Besides, we propose a method to estimate the disparity map and the camera response difference parameters simultaneously. Experiments with various real stereo images show robust performances of our algorithm.

We first review research results of camera self-calibration achieved in photogrammetry, robotics and computer vision. Then we propose a method for self-calibration of robotic hand cameras by means of active motion. Through tracking a set of world points of unknown coordinates during robot motion, the internal parameters of the cameras (including distortions), the mounting parameters as well as the coordinates of the world points are estimated. The approach is fully autonomous, in that no initial guesses of the unknown parameters are to be provided from the outside by humans for the solution of a set of nonlinear equations. Sufficient conditions for a unique solution are derived in terms of controlled motion sequences. Methods to improve accuracy and robustness are proposed by means of best model identification and motion planning. Experimental results in both a simulated and a real environments are reported.

We propose a method for self-calibration of robotic hand-mounted laser range finders by means of active motion of the robot. Through range-measuring a plane of unknown position and orientation, the mounting parameters of the range finders and the coordinates of the world plane are estimated. Systematic measurement errors can also be calibrated at the same time. The approach is fully autonomous, in that no initial guesses of the unknown parameters are to be provided from the outside by humans for the solution of a set of nonlinear equations. In fact, the initial values are all found in closed forms by the algorithm itself. Sufficient conditions for a unique solution are derived in terms of controlled motion sequences. Experimental results in both a simulated and a real environments are reported.

In this paper we propose a method for self-calibration of robotic hand cameras by means of active motion of the robot. Through tracking a set of world points of unknown coordinates, the internal parameters of the cameras (including lens distortions), the mounting parameters as well as the coordinates of the world points are estimated. The approach is fully autonomous, in that no initial guesses of the unknown parameters are to be provided from the outside by humans for the solution of a set of nonlinear equations. Sufficient conditions for a unique solution are derived in terms of controlled motion sequences. To improve robustness of the calibration, we propose to identify the best lens-distortion model by using F-test. Furthermore, a method is proposed to minimize the effect of robot motion uncertainties by motion planning. Experimental results in both a simulated and a real environments are reported.

In this paper, we propose a method for self-calibration of robotic hand-mounted laser range finders by means of active motion of the robot. Through range-measuring a plane of unknown position and orientation, the mounting parameters of the range finders and the coordinates of the world planes are estimated. Systematic measurement errors can also be calibrated at the same time. The approach is fully autonomous, in that no initial guesses of the unknown parameters are to be provided from the outside by humans for the solution of a set of nonlinear equations. In fact, the initial values are all found in closed forms by the algorithm itself. Sufficient conditions for a unique solution are derived in terms of controlled motion sequences. Experimental results in both a simulated and a real environments are reported.

Position-based visual servoing requires estimating and tracking the three dimensional position and orientation of a 3D target object from camera images. This paper describes a novel approach to the problem that consists of two steps. First, a set of spatial pose constraints is derived from image features, by means of which 3D object pose is calculated with an efficient model-fitting algorithm. Kalman-filtering is then used to estimate object velocity and acceleration. Compared to previous approaches that use Kalman-filters to directly estimate the object state from image features, the proposed method has a variety of advantages: Computation time is only O(n) rather than O(n/sup 3/) where n is the number of image features considered, sensor fusion is simplified and temporal estimation is decoupled from the choice of image features. The last point is of particular importance if occlusions that may occur during tracking are to be predicted and dynamically handled. With the tracking method proposed, a robot could be precisely controlled with respect to static objects and robustly follow targets moving in 6 degrees of freedom, while occasions were continuously predicted and appropriate features automatically selected at video rate (25 Hz). High robustness is obtained by Hough transform-based feature extraction.

This paper deals with the problem of obtaining a rough estimate of three dimensional object position and orientation from a single two dimensional camera image. Such an estimate is required by most 3-D to 2-D registration and tracking methods that can efficiently refine an initial value by numerical optimization to precisely recover 3-D pose. However the analytic computation of an initial pose guess requires the solution of an extremely complex correspondence problem that is due to the large number of topologically distinct aspects that arise when a three dimensional opaque object is imaged by a camera. Hence general analytic methods fail to achieve real-time performance and most tracking and registration systems are initialized interactively or by ad hoc heuristics. To overcome these limitations we present a novel method for approximate object pose estimation that is based on a neural net and that can easily be implemented in real-time. A modification of Kohonen's self-organizing feature map is systematically trained with computer generated object views such that it responds to a preprocessed image with one or more sets of object orientation parameters. The key idea proposed here is to choose network topology in accordance with the representation of 3-D orientation. Experimental results from both simulated and real images demonstrate that a pose estimate within the accuracy requirements can be found in more than 81% of all cases. The current implementation operates at 10 Hz on real world images.

In the near future, more and more robots will be used for servicing tasks, tasks in hazardous environments or space applications. Dextrous hands are a powerful and flexible tool to interact with these real world environments that are not specially tailored for robots. In order to grasp and manipulate real world objects, grasp planning systems are required. Grasp planning for general 3D objects is quite a complex problem requiring a large amount of computing time. Fast algorithms are required to integrate grasp planners in online planning systems for robots. This paper presents an heuristic approach towards fast planning of precision grasps for realistic, arbitrarily shaped 3D objects. In this approach a number of feasible grasp candidates are generated heuristically. These grasp candidates are qualified using an efficiently computable grasp quality measure and the best candidate is chosen. It is shown that only a relatively small number of grasp candidates has to be generated in order to obtain a good-although not optimal-grasp.

Global path planning for robot manipulators is a main topic in robot motion planning. Beyond the global character of path planners using an explicit configuration-space (C-space) representation, they demonstrate resolution completeness. Keeping the C-space resolution reasonable with growing number of degrees of freedom (DOF) of a robot is a challenging problem, since the C-space grows exponentially with the number of DOF. The presented approach leads to a high resolution representation of collision-free C-space regions and a low one of colliding C-space regions. This C-space representation by a nonequidistant C-space grid is achieved by using Kohonen's self organizing feature maps. The effectiveness of the approach is demonstrated by examples of up to 5-dimensional C-spaces used by a path planner.

We present a new method of shape from shading by using radial basis functions to parameterize the object depth. The radial basis functions are deformed by adjusting their centers, widths, and weights such that the intensity errors are minimized. The initial centers and widths are arranged hierarchically to speed up convergence and to stabilize the solution. Although the smoothness constraint is used, it can be eventually dropped out without causing instabilities in the solution. An important feature of our parametric shape-from-shading method is that it offers a unified framework for integration of multiple sensory information. We show that knowledge about surface depth and/or surface normals anywhere in the image can be easily incorporated into the shape from shading process. It is further demonstrated that even qualitative knowledge can be used in shape from shading to improve 3D reconstruction. Experimental comparisons of our method with several existing ones are made by using both synthetic and real images. Results show that our solution is more accurate than the others.

A new algorithm for registering a 3D model to a single or a sequence of 2D images without a-priori knowledge of correspondences is presented. The algorithm is composed of two iterated steps: the first step determines correspondences between image features and model data, while the second one computes the 3D rigid body transformation that minimizes the displacement of the matched points. As camera images do not yield full 3D information, the key idea proposed is to exploit the inverse perspective, i.e. the projection ray determined by the focal point of the camera and an image feature point. The inverse perspective yields the constraints for both correspondence search and pose estimation in 3D space rather than in the image plane. This not only simplifies the motion parameter estimation but also reduces matching ambiguities due to occlusions. In addition, a robust M-estimation technique reduces the impact of false correspondences. Experimental results show that complex, 3D CAD-models can be registered efficiently and accurately to images, even if image features are incomplete, fragmented and noisy.

The paper applies a previously presented method for accurate tracking of paths to force control. This approach is very simple since it does not require a joint torque/motor current interface but only a positional interface. It can be applied with elastic end-effectors (sensors) as well as with stiff environments where most elasticity is in the robot joints. In both cases deviations from the desired forces are transferred to positional deviations on joint level. The resulting path can then be controlled with high accuracy by a learned feedforward controller including the influence of the forces. The approach can be applied to the sensing of a contour or to the tracking of a known contour with high speed.

In this paper we discuss the relation between the two approaches to velocity command generation for nonredundant robotic mechanisms, which the two groups of the authors proposed recently and independently of each other. It will be shown analytically that the singularity-consistent null space based approach, and the split Jacobian approach, are equivalent. Analysis of the behavior at a singularity will be presented from the viewpoint of both approaches. An analytical example will be used to demonstrate the theoretical results.

We present a resolution complete point to point path planner for up to 6 degrees of freedom (DOF) robot manipulators, based on a discretized configuration space (C-space) representation. The robot operates in an initially known workspace (W-space). Vision integration in an easy way makes the planner capable of dealing with initially unknown obstacles. Furthermore, a Kohonen map based reorganization of the C-space increases the resolution in collision free regions of the C-space and enables path finding even for difficult paths. Experiments with low on-line computation times (lying in the order of a fraction of a second to a few seconds) demonstrate the effectiveness of the planner.

We are proposing a supervised learning approach in robot force control which enables robot programming by human demonstration. To demonstrate the task in an intuitive way, we designed a teach device which acquires the sensorimotion of the human. The design and use of the teaching device for skill acquisition is discussed. Recording of compliant motion, of the human is shown for the "put block in a corner of a box problem" and the "peg in hole problem". Skill representation is demonstrated for the first of the two problems. Finally we are describing other useful applications of the teach device in the field of world modeling and robotics.

The multilayer feedforward network has often been used for learning a nonlinear mapping based on a set of examples of the input-output data. In this paper, we present a novel use of the network, in which the example data are not explicitly given. We consider the problem of shape from shading in computer vision, where the input (image coordinates) and the output (surface depth) satisfy only a known differential equation. We use the feedforward network as a parametric representation of the object surface and reformulate the shape from shading problem as the minimization of an error function over the network weights. The stochastic gradient and conjugate gradient methods are used for the minimization. Boundary conditions for either surface depth or surface normal (or both) can be imposed by adjusting the same network at different levels. It is further shown that the light source direction can be estimated, based on an initial guess, by integrating the source estimation with the surface estimation. Extensions of the method to a wider class of problems are discussed. The efficiency of the method is verified by examples of both synthetic and real images.

This paper presents detailed description of a laboratory simulator that uses two robots fixed on earth to simulate the motion of a robot in space relative to a free-flying target. This simulator is appealing because it creates an environment on earth similar to what an astronaut observes watching from the space shuttle. Therefore, it could be a valuable tool for training of astronauts on earth. Also, this simulator uses off-the-shelf industrial robots in contrast to laboratory facilities today that use special designs of robots to achieve this same task. The proposed simulator can be valuable for studying real-time motion in space and in evaluating the effectiveness of onboard sensors and control systems. The objectives of this paper are to: i) describe the laboratory setup, ii) present the computational details of this simulator, iii) estimate the computation needs of this simulator, and iv) present a computer implementation of this simulator using two KUKA robots.

This paper presents a novel approach to singularity treatment of non-redundant robots. In particular it tries to yield a complete solution to the singularity problem. Based on the symbolic deduction of the inverse Jacobian a necessary condition is derived to discriminate between executable and non-executable motions. In this context the redundant motions in a singular robot configuration are examined and illustrated. The robot can move out of the singular configuration without any position or orientation error. Two strategies are derived to move through the singular configuration. The results although general-are verified using the six axis ROTEX robot, the first remotely controlled space robot.

In many robotics applications, e.g. telerobotics under long time delay, building a geometric world model from multisensory data is a crucial requirement. Using a sensor-in-hand configuration to generate a model referenced to the robot base, imposes specific problems. Moving the sensor frame on an arbitrary path over and around an unknown object generates a completely unordered 'data cloud'. A surface reconstruction algorithm, which is based on Kohonen's self-organizing feature maps is used to process this data cloud and to generate a useful surface description. This surface can be used for object recognition and pose estimation using algorithms which were developed in the field of range image understanding.

In April 1993 for the first time in the history of space flight, a small multisensory robot performed a number of prototype tasks on-board a spacecraft (spacelab D2 on shuttle COLUMBIA) in different operational modes that are feasible today, namely preprogrammed remotely controlled operations by the astronauts using a control ball and a stereo TV-monitor, as well as remotely controlled from ground via the human operator and machine intelligence. In these operational modes the robot successfully closed and opened connector plugs (bayonet closure), assembled structures from single parts and captured a free-floating object. Several key technologies have made this space robot technology experiment ROTEX a big success: multisensory gripper technology, local (shared autonomy) sensory feedback control concepts, and the powerful delay-compensating 3D-graphics simulation (predictive simulation) in the telerobotic ground station. This paper focusses on the tele-sensor-programming approach and the predictive simulation used for remote ground control.>

In this paper a method is presented to reconstruct and model an unknown three-dimensional surface which is described by an unordered cloud of sampled surface points. For that purpose Kohonen's self-organizing feature map is modified accordingly. This algorithm models the 2-D subspace in the 3-D input space by defining the appropriate parameter grid. In the second part of the paper the authors discuss the extensions of Kohonen's algorithm, which were necessary to handle multiple sensor input, addressing orientation discontinuities and defining the reconstruction resolution according to surface properties. The result is a method which can perform satisfactorily on sparse as well as on dense input data. Because the surface is described in a parameterized form viewpoint independence is inherent. In the last part experiments with real and simulated data are presented, in which the ROTEX telerobotic station served as background scenario.>

We propose the so-called TeleSensor programming concept that uses sensory perception to achieve local autonomy in robotic manipulation. Sensor based robot tasks are used to define elemental moves within a high level programming environment. This approach is applicable in both, the real robot's world and the simulated one. Beside the graphical off-line programming concept, the range of application lies especially in the field of teleoperation with large time delays. A shared autonomy concept is proposed that distributes intelligence between man and machine. The feasibility of graphically simulating the robot within its environment is extended by emulating different sensor functions to achieve a correct copy of the real system behaviour as far as possible. The programming paradigm is supported by a sophisticated graphical man machine interface. Sensor fusion aspects with respect to autonomous sensor controlled task execution are discussed as well as the interaction between the real and the simulated system.>

Advances in visual servoing theory and practice now make it possible to accurately and robustly position a robot manipulator relative to a target. Both the vision and control algorithms are extremely simple, however they must be initialized on task-relevant features in order to be applied. Consequently, they are particularly well-suited to telerobotics systems where an operator can initialize the system but round-trip delay prohibits direct operator feedback during motion. This paper describes the basic theory behind feature-based visual servoing, and discusses the issues involved in integrating visual servoing into the ROTEX space teleoperation system.>

The paper discusses delay-compensating techniques when operating a space robot from ground or from a another remote spacecraft. These kind of techniques have been a key element in the space robot technology experiment ROTEX that has successfully flown with shuttle COLUMBIA end of April 93. During this "spacelab-D2"-mission for the first time in the history of space flight a small, multisensory robot (i.e. provided with modest local intelligence) has performed prototype tasks on board a spacecraft in different operational modes, namely preprogrammed (and reprogrammed from ground), remotely controlled (teleoperated) by the astronauts, but also remotely controlled from ground via the human operator as well as via machine intelligence. In these operational modes the robot successfully closed and opened connector plugs (bayonet closure), assembled structures from single parts and captured a free-floating object. This paper focuses on the powerful delay-compensating 3D-graphics simulation (predictive simulation) concepts that were realized in the telerobotic ground station and which allowed the authors to compensate delays of up to 7 sec e.g. when grasping the floating object fully automatically from ground.>

A learning method is presented which improves the dynamic accuracy of conventional industrial robots with integrated position control. The method is based on feedforward control being able to follow off-line programmed trajectories with high speed and negligible pose errors. For learning, the robot has to be moved along a given path. The algorithm then estimates a simple model. This model is used to build a controller which is able to modify positional commands, thus reducing the positional path error from some millimeters to approximately 0.2 mm for a Manutec r2 robot. This improvement is valid also for other, non-trained trajectories. For repetitive control of a single path the error is even lower. Measurements of path accuracy are verified using data of a force/torque sensor during tracking a known contour.>

We present a path planning algorithm for robot manipulators working in an initially known environment. An efficient method constructs the configuration space (C-space) of the robot and expands a local-minima-free numerical potential field in it. So the robot is able to move very fast and collision-free within its workspace (W-space). An extension of that principle allows the robot to handle autonomously new obstacles appearing in its W-space in a fast manner. We have implemented the planner on our 6 degrees of freedom (DOF)-considering 5 of them-space robot ROTEX working within a tight workcell. In order to reduce the off-line computation time of the C-space, the C-space may be only partially constructed taking approximately 5 minutes on a Silicon Graphics IRIS Indigo R4000. Pathfinding solutions for many practical situations are then produced in less than 1 minute.>

Outlines key technologies in the approach of the author's research establishment to space robotics. Based on multisensory gripper technology, local on-board sensory feedback, and predictive graphic simulation (with emphasis on sensory simulation) a tele-sensor programming concept is introduced that allows sensor-based teleoperation in spite of large signal delays as well as sensor-based off-line programming following a "learning by showing" concept. A small multisensory robot based on these concepts has flown in space with a ten-day Space Shuttle mission. This robot technology experiment ROTEX was very successful and showed that, with these sensor-based concepts, even present-day space robots can perform different prototype tasks in a variety of operational modes, including automatic (reprogrammable) operation, and on-board teleoperation using human and/or machine intelligence.

In early 1993 the space robot technology experiment ROTEX was flown with space-shuttle Columbia. A multisensory robot onboard the spacecraft successfully worked in autonomous modes, teleoperated by astronauts, as well as in different telerobotic ground control modes. These included online teleoperational and telesensor-programming: a task-level oriented programming technique involving learning-by-showing concepts in a virtual environment. The robot's key features were its multisensory gripper and the local sensory feedback schemes that are the basis for shared autonomy. The corresponding man-machine interface concepts, which use a six-degree-of-freedom non-force-reflecting control ball and visual feedback to the human operator, are explained. Stereographic simulation on the ground was used to predict not only the robot's free motion but even the sensor-based path refinement onboard. Prototype tasks performed by this space robot were the assembly of a truss structure, connecting/disconnecting an electrical plug (orbit replaceable unit exchange), and grasping free-floating objects.>

Describes a method of determining a robot end-effector's motion required to achieve a standard position and orientation relative to an object through learning. By using a back-propagation network, the authors establish the direct mapping from 'what is seen' to 'what should be done'. The method does not need camera calibration, nor hand-eye calibration, nor explicit object model. Some general rules for correct learning are presented. A recursive scheme of movement control is designed with convergence proof. The method is simulated on an application object and shows promising application potential.>

A very general three-level learning method for self-improvement of the parameters of a force feedback controller is demonstrated in contour tracking tasks. It is assumed that no model is known a priori, either of the robot or of the contour to be tracked. The system identifies such a model, including information about its reliability. The model and estimated noise were used to generate optimal control actions for the sample trajectory. They were then used for estimation of the parameters of the controller. This controller then produces a new trajectory, which in turn could be optimized and trained. Kalman filter techniques were applied in all adaptation levels involved. Learning was possible off-line or online. The model and controller may be based on linear difference equations or include nonlinear mappings as associative or tabular memories or neural networks. It was shown that even for a linear controller substantial improvements could be attained as no assumptions were needed about the bandwidth.>

The space robot technology experiment ROTEX to fly with the next spacelab mission D2 in January 1993 provides a multisensory robot which is supposed to work in an autonomous mode, teleoperated by astronauts, as well as in different telerobotic ground control modes.These include on-line teleoperation and telesensorprogramming, a task-level oriented programming technique involving `learning by showing'concepts in a virtual environment. The robot's key features are its multisensory gripper and the local sensory feedback schemes which are the basis for shared autonomy. The corresponding man-machine interface concepts using a 6 dof non- force-reflecting control ball and visual feedback to the human operator are explained.Stereographic simulation on ground is supposed to predict even the sensor based path refinement on board, while realtime fusion of stereo images and laser range information helps to predict the motion of floating objects to be grapsed from ground, too. Other prototype tasks to be performed by this space robot are the assembly of a truss structure and connecting/ disconnecting an electrical plug (orbit replaceable unit exchange ORU).

A robotic vision system for grasping a free-flying polyhedron in space has been developed using stereo vision and laser range finders. Real-time motion estimation and sensor fusion is achieved by using prior knowledge of the object. A maximum likelihood parameter estimation is developed for rotational symmetric polyhedrons, and the 3D transformations for fusing many different sensors into one coordinate frame are given. The minimization is solved using the sequential quadratic programming technique, which has proved to be a robust and efficient method. Included are simulation results performed on the hardware that will be used in the ROTEX space robot technology experiment.>

The method of Fourier descriptors is extended to produce a set of normalized coefficients which are invariant under any affine transformation (translation, rotation, scaling, and shearing). The method is based on a parameterized boundary description which is transformed to the Fourier domain and normalized there to eliminate dependencies on the affine transformation and on the starting point. Invariance to affine transforms allows considerable robustness when applied to images of objects which rotate in all three dimensions, as is demonstrated by processing silhouettes of aircraft maneuvering in three-space.>

The problems that arise when sensor-controlled robots in space are teleoperated from ground stations are discussed. A supervisory control concept is described that makes it possible to realize shared control between teleoperator and sensor-controlled robot in a variety of configurations. Predictive 3D computer graphics currently seems to be the only way to cope successfully with the problem of transmission-time delays of several seconds. Appropriate estimation schemes in combination with knowledge-based world modeling are outlined, which include models of the delay lines, the robot, moving objects, etc., and which derive the necessary updates from sensory data as they are sent down from the spacecraft to earth (e.g. via real-time stereo vision). The space robot technology experiment Rotex scheduled for the next German Spacelab mission (2D) is taken as a basis for the problem description.>

A discussion is presented of problems of multisensory integration and feedback in a robot gripper. Various contact and noncontact sensors are described and it is shown how a number of such 'low-dimensional' sensors are integrable in an advanced robot gripper. The resulting gripper uses only a few data and power supply connections out of the wrist. Feedback structures are outlined with the main emphasis on different alternatives in force-torque feedback; the various approaches proposed in the literature are considered in more detail. The problem of sensor data fusion, including the redundancy problem, is briefly explained.>

The Paper outlines the concepts of a robot technology experiment ROTEX we have proposed to fly with the next German spacelab mission D2 (originally planned for 88, now delayed for at least two years). It provides a small, six axis robot inside a space-lab rack, equipped with a multisensory gripper (force/torque, an array of range finders, stereo optical fibers). The robot is supposed to handle a biological experiment, to perform several assembly and "servicing tasks" and to grasp floating objects. The paper focusses on the man-machine and supervisory control concepts for teleoperation from the spacecraft and from ground and expecially explains the predictive estimation schemes for an extensive use of delay-compensating 3D-computer graphics.

The paper first outlines the recent sensory developments in the DFVLR robotics lab, these are different force-torque sensors, laser range finders, inductive sensors and sensor balls for robot and 3D-computer-grafic control. With the example of our proposed multisensory arrangement implying vision, range sensing, force-torque and speech, the fine-motion planning and path generation techniques as developed in our lab are discussed. The special case of a two-arm cooperative robot using two force-torque-sensors is treated in more detail. Practical results are demonstrated in a film including cooperative two arm robot control.

A concept for sensory feedback and sensor-based teach-in in robotics is presented. Based on previous work of different authors a scheme of hybrid sensory-position control is proposed and has been realized that interpretes every "rudimentary" teach command in terms of sensory interaction with the environment by generating the socalled "artificial constraints" and the refined path automatically. Robots with arbitrarily programmable stiffness are one outcome of our technique. Motion commands and sensor data are stored together. The latter ones are then available as reference values for the repetition mode in a possibly changing environment. It is shown that by introducing pseudo-forces/torques the proposed techniques are equally applicable to different kinds of sensor, as are force-torque-sensors, range finders or inductive sensors. The "sensor-ball"-technique as developed at DFVLR is discussed as one physical realization. Operational systems of this kind, first tested with an ASEA robot in our lab, are going into industrial application now.