Quadratic Programs (QPs) are widely used in the control of walking robots, especially in Model Predictive Control (MPC) and Whole-Body Control (WBC). In both cases, the controller design requires the formulation of a QP and the selection of a suitable QP solver, both requiring considerable time and expertise. While computational performance benchmarks exist for QP solvers, studies comparing optimal combinations of computational hardware (HW), QP formulation, and solver performance are lacking. In this work, we compare dense and sparse QP formulations, and multiple solving methods on different HW architectures, focusing on their computational efficiency in dynamic walking of four-legged robots using MPC. We introduce the Solve Frequency per Watt (SFPW) as a performance measure to enable a cross-hardware comparison of the efficiency of QP solvers. We also benchmark different QP solvers for WBC that we use for trajectory stabilization in quadrupedal walking. As a result, this paper recommends a starting point for practitioners on the selection of QP formulations and solvers for different HW architectures in walking robots and indicates which problems should be devoted the greater technical effort.

The employment of walking excavator robots, endowed with hybrid locomotion capabilities, holds considerable promise in facilitating the execution of intricate tasks in challenging terrain environments. A critical skill for such a system pertains to traversing obstacles through stepping locomotion, a process entailing the momentary disengagement of the end-effectors from the ground. Existing solutions are encumbered by two significant limitations. Primarily, they are often too cumbersome to develop and implement due to the complexity of the problem formulations. Secondly, they present restrictions on the available avenues to influence the behavior, hindering the effective leveraging of domain knowledge to achieve the intended objective.This research proposes an alternative approach to learning the stepping locomotion. The proposed method employs a hierarchical reinforcement learning strategy, wherein the complex control task is decomposed into multiple subtasks, each aligned with a sub-objective defined as a reward function. The training of these subtasks is conducted individually, starting from the lowest level and progressing to the higher levels, predominantly utilizing deep reinforcement learning. Additionally, the masking of invalid actions is utilized to guide the controller during training, offering enhanced opportunities to influence behavior while using only simple formulations. Notably, the proposed approach has been successfully trained for three distinct stepping scenarios: obstacle, step, and gap, underscoring the versatility of the controller. The design of the controller, along with the results of training and evaluation in simulation, is presented herein.

Many real-world applications require legged robots to be able to carry variable payloads. Model-Based controllers such as model predictive control (MPC) have become the de facto standard in research for controlling these systems. However, most model-based control architectures use fixed plant models, which limits their applicability to different tasks. In this paper, we present a Kalman filter (KF) formulation for online identification of the mass and center of mass (COM) of a four-legged robot. We evaluate our method on a quadrupedal robot carrying various payloads and find that it is more robust to strong measurement noise than classical recursive least squares (RLS) methods. Moreover, it improves the tracking performance of the model-based controller with varying payloads when the model parameters are adjusted at runtime.

In robotics, structural design and behavior optimization have long been considered separate processes, resulting in the development of systems with limited capabilities. Recently, co-design methods have gained popularity, where bi-level formulations are used to simultaneously optimize the robot design and behavior for specific tasks. However, most implementations assume a serial or tree-type model of the robot, overlooking the fact that many robot platforms incorporate parallel mechanisms. In this paper, we present a first co-design formulation that explicitly incorporates parallel coupling constraints into the dynamic model of the robot. In this framework, an outer optimization loop focuses on the design parameters, in our case the transmission ratios of a parallel belt-driven manipulator, which map the desired torques from the joint space to the actuation space. An inner loop performs trajectory optimization in the actuation space, thus exploiting the entire dynamic range of the manipulator. We compare the proposed method with a conventional co-design approach based on a simplified tree-type model. By taking advantage of the actuation space representation, our approach leads to a significant increase in dynamic payload capacity compared to the conventional co-design implementation.

Real-world robotic applications, from autonomous exploration to assistive technologies, require adaptive, interpretable, and data-efficient learning paradigms. While deep learning architectures and foundation models have driven significant advances in diverse robotic applications, they remain limited in their ability to operate efficiently and reliably in unknown and dynamic environments. In this position paper, we critically assess these limitations and introduce a conceptual framework for combining data-driven learning with deliberate, structured reasoning. Specifically, we propose leveraging differentiable physics for efficient world modeling, Bayesian inference for uncertainty-aware decision-making, and meta-learning for rapid adaptation to new tasks. By embedding physical symbolic reasoning within neural models, robots could generalize beyond their training data, reason about novel situations, and continuously expand their knowledge. We argue that such hybrid neuro-symbolic architectures are essential for the next generation of autonomous systems, and to this end, we provide a research roadmap to guide and accelerate their development.

Optimal behaviours of a system to perform a specific task can be achieved by leveraging the coupling between trajectory optimization, stabilization, and design optimization. This approach is particularly advantageous for underactuated systems, which are systems that have fewer actuators than degrees of freedom and thus require for more elaborate control systems. This paper proposes a novel co-design algorithm, namely Robust Trajectory Control with Design optimization (RTC-D). An inner optimization layer (RTC) simultaneously performs direct transcription (DIRTRAN) to find a nominal trajectory while computing optimal hyperparameters for a stabilizing time-varying linear quadratic regulator (TVLQR). RTC-D augments RTC with a design optimization layer, maximizing the system’s robustness through a time-varying Lyapunov-based region of attraction (ROA) analysis. This analysis provides a formal guarantee of stability for a set of off-nominal states. The proposed algorithm has been tested on two different underactuated systems: the torque-limited simple pendulum and the cart-pole. Extensive simulations of off-nominal initial conditions demonstrate improved robustness, while real-system experiments show increased insensitivity to torque disturbances.

This paper presents the design, analysis, and performance evaluation of RicMonk, a novel three-link brachiation robot equipped with passive hook-shaped grippers. Brachiation, an agile and energy-efficient mode of locomotion observed in primates, has inspired the development of RicMonk to explore versatile locomotion and maneuvers on ladder-like structures. The robot’s anatomical resemblance to gibbons and the integration of a tail mechanism for energy injection contribute to its unique capabilities. The paper discusses the use of the Direct Collocation methodology for optimizing trajectories for the robot’s dynamic behaviors and stabilization of these trajectories using a Time-varying Linear Quadratic Regulator. With RicMonk we demonstrate bidirectional brachiation, and provide comparative analysis with its predecessor, AcroMonk - a two-link brachiation robot, to demonstrate that the presence of a passive tail helps improve energy efficiency. The system design, controllers, and software implementation are publicly available on GitHub at https://github.com/dfki-ric-underactuated-lab/ricmonk and the video demonstration of the experiments can be viewed at https://youtu.be/hOuDQI7CD8w.

Exploration missions on extra terrestrial celestial bodies are to date performed by complex and heavy robotic systems. The trend is towards lighter modular systems that can be (re)configured in situ according to mission specific requirements. To facilitate flexible configurability, a multifunctional interconnect is used to mechanically couple the involved systems while providing electrical power and data transmission. The paper presents the further development of the reliable electromechanical interface (EMI) from the TransTerrA project, which has been proven in several field tests and reached TRL 4. Docking under loads of up to 550 N has been successfully tested with the new design. The experiments presented include undocking at various inclinations with different loads expected for the application scenario. The maximum determined static load that can be carried by the further developed EMI is 2000 N. In further experiments, new contact blocks responsible for the transfer of electrical power and data were tested for water resistance and resilience to environmental factors, as well as power and data transfer. The obtained results will be helpful in the development of a multi-functional interface suitable for lunar applications and missions having similar challenging environmental conditions.

In this paper, we present an attitude control scheme for an autonomous underwater vehicle (AUV), which is based on incremental nonlinear dynamic inversion (INDI). Conventional model-based controllers depend on an exact model of the controlled system, which is difficult to find, especially for marine vehicles subject to highly nonlinear hydrodynamic effects. INDI trades off model accuracy with sensor accuracy by incorporating acceleration feedback and actuator output feedback to linearize a nonlinear system incrementally. Existing research primarily focuses on studying INDI on unmanned aerial vehicles. However, there is barely any research on controlling marine vehicles using INDI. The control task we are performing is a 90 degrees pitch-up maneuver, where the dual-arm intervention AUV Cuttlefish transitions from a horizontal traveling pose to a vertical intervention pose. We compare INDI to a classical model-based control scheme in the maritime test basin at DFKI RIC, Germany, and we find that INDI keeps the AUV much more steady both in the transitioning phase as well as in the station keeping phase.

In real-world scenarios, robots come closer to humans in many applications, sharing the same workspace or even manipulating the same objects. To ensure safe and intuitive collaboration, it is crucial to have an accurate knowledge of the human’s 3D position in space, which should be estimated with high precision, high frequency and low latency. However, individual sensors such as inertial measurement units (IMUs) or cameras cannot meet all requirements for reliable human pose estimation under conditions such as long operating times, large distances and occlusions. In this study, we highlight the limitations of different visual pose methods and present a fused approach for real-time estimation of the 3D position of the human pelvis using machine learning-based visual pose from a monocular camera and an IMU sensor suit. The multimodal fusion is based on the Invariant Extended Kalman filter (InEKF) on Lie Groups, which fuses drift-free visual poses with high-frequency inertial measurements in a loosely-coupled manner. The evaluation is performed on a recorded dataset of multiple subjects performing various experimental scenarios. The results show that the fused approach can increase the accuracy and robustness of the estimates, taking a step closer towards smooth human-robot collaboration.

Trajectory optimization methods have become ubiquitous for the motion planning and control of underactuated robots for e.g., quadrupeds, humanoids etc. While they have been extensively used in the case of serial or tree type robots, they are seldomly used for planning and control of robots with closed loops. Series-parallel hybrid topology is quite commonly used in the design of humanoid robots, but they are often neglected during trajectory optimization and the movements are computed for a serial abstraction of the system and then the solution is mapped to the actuator coordinates. As a consequence, the full capability of the robot cannot be exploited. This paper presents a case study of trajectory optimization for series-parallel hybrid robot by taking into account all the holonomic constraints imposed by the closed kinematic loops present in the system. We demonstrate the advantages of this consideration with a weightlifting task on RH5 Manus humanoid in both simulation and experiments.

This work proposes the use of conditional Generative Adversarial Networks (cGANs) for acoustic-based 3D reconstruction. Acoustics being the most reliable sensor modality in underwater domains is accompanied with the loss of elevation angle in its images. The challenge of recovering the missing dimension in acoustic images have pushed researchers to try various methods and approaches over the past years. cGANs being an image-to-image translation method makes it possible to learn a desired style, and transforms the data from one modality to another. This was applied here as a way of transforming an acoustic image into another form which contains the elevation characteristics, such as depth images. Depth images are hard to acquire underwater, thus data was generated synthetically and used for training and testing the deep learning model. As a way of performance enhancement, real data was collected for training a Cycle-GAN network in the aim of transferring the realistic style into the synthetically generated images. Simulation experiments were conducted to evaluate the system and find out the best experimental setup, which was then used to carry out the real experiment. The system performed dense 3D reconstruction of the scanned object and proved to be applicable in real environments.

Legged locomotion is arguably the most suited and versatile mode to deal with natural or unstructured terrains. Intensive research into dynamic walking and running controllers has recently yielded great advances, both in the optimal control and reinforcement learning (RL) literature. Hopping is a challenging dynamic task involving a flight phase and has the potential to increase the traversability of legged robots. Model based control for hopping typically relies on accurate detection of different jump phases, such as lift-off or touch down, and using different controllers for each phase. In this paper, we present a end-to-end RL based torque controller that learns to implicitly detect the relevant jump phases, removing the need to provide manual heuristics for state detection. We also extend a method for simulation to reality transfer of the learned controller to contact rich dynamic tasks, resulting in successful deployment on the robot after training without parameter tuning.

Human-Robot Interaction (HRI) becomes more and more important in a world where robots integrate fast in all aspects of our lives but HRI applications depend massively on the utilized robotic system as well as the deployment environment and cultural differences. Because of these variable dependencies it is often not feasible to use a data-driven approach to train a model for human intent recognition. Expert systems have been proven to close this gap very efficiently. Furthermore, it is important to support understandability in HRI systems to establish trust in the system. To address the above-mentioned challenges in HRI we present an adaptable python library in which current state-of-the-art Models for context recognition can be integrated. For Context-Based Intention Recognition a two-layer Bayesian Network (BN) is used. The bayesian approach offers explainability and clarity in the creation of scenarios and is easily extendable with more modalities. Additionally, it can be used as an expert system if no data is available but can as well be fine-tuned when data becomes available

In this paper, we propose a methodology for co-optimizing application specific neural network (NN) accelerators for accuracy and energy expenditure per inference. The architecture of the NN is co-optimized with the concrete ASIC implementation of the accelerator to provide reliable estimates of the energy efficiency. While not constrained to a specific application or NN accelerator architecture, the method is demonstrated on an application specific NN accelerator for the detection of atrial fibrillation in human electrocardiograms that is implemented in 22FDX/FDSOI technology. The NN accelerator is highly parameterizable, i.e., it can map NNs with different architectural properties to a synthesizeable register transfer level representation. The parameter space is further expanded by the parameters of the physical implementation (e.g. memories, clocking, voltage domains). Since the evaluation of accuracy and energy efficiency for a specific parameter combination is computationally expensive, different hyperparameter optimization methods are used and evaluated, including Bayesian Optimization, which tries to find the optimal neural network architecture and physical implementation parameters with a minimum number of training, simulation and evaluation steps.

In this work, we investigate the influence of labeling methods on the classification of human movements on data recorded using a marker-based motion capture system. The dataset is labeled using two different approaches, one based on video data of the movements, the other based on the movement trajectories recorded using the motion capture system. The dataset is labeled using two different approaches, one based on video data of the movements, the other based on the movement trajectories recorded using the motion capture system. The data was recorded from one participant performing a stacking scenario comprising simple arm movements at three different speeds (slow, normal, fast). Machine learning algorithms that include k-Nearest Neighbor, Random Forest, Extreme Gradient Boosting classifier, Convolutional Neural networks (CNN), Long Short-Term Memory networks (LSTM), and a combination of CNN-LSTM networks are compared on their performance in recognition of these arm movements. The models were trained on actions performed on slow and normal speed movements segments and generalized on actions consisting of fast-paced human movement. It was observed that all the models trained on normal-paced data labeled using trajectories have almost 20% improvement in accuracy on test data in comparison to the models trained on data labeled using videos of the performed experiments.

It is well established that a stiff structure along with an optimal mass distribution are key features to perform dynamic movements, and parallel designs provide these characteristics to a robot. This work presents the new upper-body design of the humanoid robot RH5 named RH5 Manus with series-parallel hybrid design. The new design choices allow us to perform dynamic motions including tasks that involve a payload of 4 kg in each hand and fast boxing motions. The parallel kinematics combined with an overall serial chain of the robot provides us with high force production along with a larger range of motion and low peripheral inertia. The robot is equipped with backdrivable actuators with current sensing, force-torque sensors, stereo camera, laser scanners, high-resolution encoders etc that provide interaction with operators and environment. We generate several diverse dynamic motions using trajectory optimization, and successfully execute them on the robot with accurate trajectory and velocity tracking, while respecting joint rotation, velocity, and torque limits.

Parallel mechanisms are becoming increasingly popular as subsystems in various robots due to their superior stiffness, payload-to-weight ratio, and dynamic properties. The serial connection of parallel subsystems leads to series-parallel hybrid robots, which are more difficult to model and control than serial or tree-type systems. At the same time, Whole-Body Control (WBC) has become the method of choice in the control of robots with redundant degrees of freedom, e.g., legged robots. However, most state-of-the-art WBC frameworks can only deal with serial or tree-type robot topologies. In this paper, we describe a computationally efficient framework for Whole-Body Control of series-parallel hybrid robots subjected to a large number of holonomic constraints. In contrast to existing WBC frameworks, our approach describes the optimization problem in the actuation space of a series-parallel robot, which provides better exploitation of the feasible workspace, higher accuracy, and more transparent behavior near singularities. We evaluate the proposed framework on two different humanoids with series-parallel architecture and compare its performance to a WBC approach for tree-type robots.

Musical dancing is an ubiquitous phenomenon in the human society. Providing robots the ability to dance has the potential to make the human robot co-existence more acceptable in our society. Hence, dancing robots have generated a considerable research interest in the recent years. In this paper, we present a novel formalization of robot dancing as planning and control of optimally timed actions based on beat timings and additional features extracted from the music. We showcase the use of this formulation in three different variations: with input of human expert choreography, imitation of a predefined choreography, and automated generation of a novel choreography. Our method has been validated on four different musical pieces, both in simulation and on a real robot, using the upper-body humanoid robot RH5 Manus.

Modeling closed loop mechanisms is a necessity for the control and simulation of various systems and poses a great challenge to rigid body dynamics algorithms. Solving the forward and inverse dynamics for such systems require resolution of loop closure constraints which are often solved via numerical procedures. This brings an additional burden to these algorithms as they have to stabilize and control the loop closure errors. In order to avoid this issue, analytical solutions are preferred for commonly studied parallel mechanisms. This paper has two contributions. Firstly, it reports a case study on a modular and hybrid numerical-analytical approach to model and control series-parallel hybrid robots which are subjected to large number of holonomic constraints. The approach exploits the modularity in the robot design to combine analytical loop closure for the known submechanisms and numerical loop closure for submechanisms where analytical solutions are not available. This offers an edge over purely numerical approach in terms of computational efficiency. Secondly, an adaption of the constraint embedding approach in Articulated Body Algorithm (ABA) is presented which yields a recursive algorithm in minimal coordinates for computing the forward dynamics of series-parallel hybrid systems. The proposed modification exploits the Lie group formulations and allows easy implementation of recursive forward dynamics of constrained systems in state of the art multi-body solvers.

Unlike fully actuated systems, the control of underactuated robots necessitates the use of passive dynamics to fulfill control objectives. Hence, there is an increased interdependence between their design parameters and the closed loop performance. This paper proposes a novel approach for co-optimization of robot design and controller parameters for increased certifiable stability obtained with means of region of attraction analysis and gradient free optimization. In particular, it discusses the co-optimization problem of a gymnastic acrobot robot where the design and the controller are optimized to have a large region of attraction (ROA) taking into account the closed loop dynamics of the non-linear system stabilized by a linear quadratic regulator (LQR) controller. The results are validated by extensive simulation of the acrobot's closed loop dynamics.

Flexible adaptation of learning strategy depending on online changes of the user's current intents have a high relevance in human-robot collaboration. In our previous study, we proposed an intrinsic interactive reinforcement learning approach for human-robot interaction, in which a robot learns his/her action strategy based on intrinsic human feedback that is generated in the human's brain as neural signature of the human's implicit evaluation of the robot's actions. Our approach has an inherent property that allows robots to adapt their behavior depending on online changes of the human's current intents. Such flexible adaptation is possible, since robot learning is updated in real time by human's online feedback. In this paper, the adaptivity of robot learning is tested on eight subjects who change their current control strategy by adding a new gesture to the previous used gestures. This paper evaluates the learning progress by analyzing learning phases (before and after adding a new gesture for control). The results show that the robot can adapt the previously learned policy depending on online changes of the user's intents. Especially, learning progress is interrelated with the classification performance of electroencephalograms (EEGs), which are used to measure the human's implicit evaluation of the robot's actions.

Learning the dynamics of robots from data can help achieve more accurate tracking controllers, or aid their navigation algorithms. However, when the actual dynamics of the robots change due to external conditions, on-line adaptation of their models is required to maintain high fidelity performance. In this work, a framework for on-line learning of robot dynamics is developed to adapt to such changes. The proposed framework employs an incremental support vector regression method to learn the model sequentially from data streams. In combination with the incremental learning, strategies for including and forgetting data are developed to obtain better generalization over the whole state space. The framework is tested in simulation and real experimental scenarios demonstrating its adaptation capabilities to changes in the robot's dynamics.

It is becoming increasingly popular to use parallel mechanisms as modular subsystem units in the design of various robots for their superior stiffness, payload-to-weight ratio and dynamic properties. This leads to series-parallel hybrid robotic systems which pose several challenges in their modeling and control e.g. resolution of loop closure constraints, large size of their spanning tree etc. These robots are typically position-controlled and when equipped with real time dynamic control, often a simplified inverse dynamic model of these systems is utilized. However, the trade-offs of this model simplification has not been studied previously. This paper presents a representative study of the neglected dynamics by introducing some error metrics which are useful in highlighting the advantages and disadvantages of such model simplification. The study is guided with the help of a series-parallel humanoid leg which has been recently developed at DFKI-RIC.

Since early in robotics the performance of odometry techniques has been of constant research for mobile robots. This is due to its direct influence on localization. The pose error grows unbounded in dead-reckoning systems and its uncertainty has negative impacts in localization and mapping (i.e. SLAM). The dead-reckoning performance in terms of residuals, i.e. the difference between the expected and the real pose state, is related to the statistical error or uncertainty in probabilistic motion models. A novel approach to model odometry errors using Gaussian processes (GPs) is presented. The methodology trains a GP on the residual between the non-linear parametric motion model and the ground truth training data. The result is a GP over odometry residuals which provides an expected value and its uncertainty in order to enhance the belief with respect to the parametric model. The localization and mapping benefits from a comprehensive GP-odometry residuals model. The approach is applied to a planetary rover in an unstructured environment. We show that our approach enhances visual SLAM by efficiently computing image frames and effectively distributing keyframes.

In this work we investigate the identification of a motion model for an autonomous underwater vehicle by applying different machine learning (ML) regression methods. By using the data collected from the robot's on-board navigation sensors, we train the regression models to learn the damping term which is regarded as one of the most uncertain components of the motion model. Four regression techniques are investigated namely, artificial neural networks, support vector machines, kernel ridge regression, and Gaussian processes regression. The performance of the identified models is tested through real experimental scenarios performed with the AUV Leng. The novelty of this work is the identification of an underwater vehicle's motion model, for the first time, through machine learning methods by using the robot's onboard sensory data. Results show that the damping model learned with nonlinear methods yield better estimates than the simplified linear and quadratic model which is identified with least-squares technique.

The work presented in this paper describes the use and evaluation of machine learning techniques like neural networks and support vector regression to learn a model of magnetic field distortions often induced in inertial measurement units using magnetometers by changing currents, postures or configurations of a robotic system. Such a model is needed in order to compensate the local dynamic distortions, especially for complex and confined robotic systems, and to achieve more robust and accurate ambient magnetic field measurements. This is crucial for a wide variety of autonomous navigation purposes from simple heading estimation over standard SLAM approaches to sophisticated magnetic field based localization techniques. The approach was evaluated in a laboratory setup and with a complex robotic system in an outdoor environment.

This paper presents the control strategies used to adapt the actively articulated suspension system of the rover SherpaTT to irregular terrain. Experimental validation of the approach with the physical system is conducted and presented. The coordinated control of the legs constituting the suspension system is encapsulated in a Ground Adaption Process (GAP) that operates independently from high level motion commands. The GAP makes use of force and orientation measurements to control the suspension system with 20 active degrees of freedom. The active suspension is used to achieve multi-objective terrain adaption encompassing (i) active force distribution at the wheel-ground contact points, (ii) keeping all wheels in permanent ground contact, and (iii) body orientation w.r.t. gravity.

A relevant issue of neuro-interfacing wearable robots in rehabilitation is the necessity to have training data, since the collection of sufficient data from patients within a reasonable recording time is not always possible. However, the use of historic data (e.g., session-to-session transfer, subject-to-subject transfer) can often lead to a reduction in classification performance which is affected by the selection of the historic data (i.e., which historic data was chosen for transfer). In this paper, we analyze two approaches to handle this reduction. First, we used incremental algorithms that can be adapted to the current session when trainable components (the spatial filter and the classifier) are transferred between different sessions. Second, we increased the number of sessions to learn more generalized models. To evaluate the approaches, we used electroencephalographic data that was recorded as training data for demonstrating our neuro-interfacing wearable robot in the application of upper-body sensorimotor rehabilitation. The data was collected from the same healthy subject on 14 different days (14 sessions). Our results showed that the use of a mixture of training sessions improved the classification performance. Further, we could show that the adaptive approaches contributed to less variability in performance that allows the system to be more robust. Hence, one can efficiently use both approaches (i.e., adapting and generalizing the models) depending on how much training data is available. Finally, the analyzed approaches are very promising to increase system applicability in upper-body sensorimotor robotic rehabilitation.

Kinematically complex robots such as legged robots provide a large degree of mobility and flexibility, but demand a sophisticated motion control, which has more tunable parameters than a general planning and decision layer should take into consideration. A lot of parameterizations exist which produce locomotion behaviors that fulfill the desired action but with varying performance, e.g., stability or efficiency. In addition, the performance of a locomotion behavior at any given time is highly depending on the current environmental context. Consequently, a complex mapping is required that closes the gap between robot-independent actions and robot-specific control parameters considering the environmental context and a given prioritization of performance indices.

This paper presents the mechanical development and testing of a docking device for a highly heterogeneous self-reconfigurable multi-module/multi-robot system. The overall system is meant to emulate a robotic lunar exploration mission. The docking device, more precisely the electro-mechanical interface (EMI), is an advancement of the reliable electromechanical connection of the project RIMRES. Since possible combinations and roles of modules in the multi-robot system are defined before a mission, a gender-principle approach with one active and one passive face to be mated was chosen. The experiments in this paper are conducted to compare the improved mechanical design with the previous design. With the new design, docking is successfully tested under loads up to 800 N. The experiments presented include attaching and detaching in different EMI orientations with various loads, exceeding those expected for the application scenario. In further experiments operations under heavy dust/small particle contamination are presented.

The estimation of robot's motion at the prediction step of any localization framework is commonly performed using a motion model in conjunction with inertial measurements. In the context of field robotics, articulated mobile robots have complex chassis. They might require a complete model in comparison with the traditionally used planar assumption. In this paper, we use a Jacobian motion model-based approach for real-time inertial-aided odometry. The work makes use of the transformation approach [1] to accurately model 6-DoF kinematics. The algorithm relates normal forces with the probability of a contact-point to slip. The result increases the accuracy by weighting the least-squares solution using static forces prediction. The method is applied to the Asguard v3 system, a simple but highly capable leg-wheel hybrid robot. The performance of the approach is demonstrated in extensive field testing within different unstructured environments. In-depth error analysis and comparison with planar odometry is discussed, resulting in a more accurate localization.

The work presented in this paper is part of the RIMRES1project. We describe the design and development of an electromechanical interface for combining heterogeneous modules. The interface has a male and a female face and allows docking in 90-degree steps. The developed concept guarantees a secure connecting and disconnecting in rough environments with fine dust as existing on celestial bodies such as Mars and Moon. A short introduction into the project RIMRES is given with focus on the modularity of the system. After providing the design considerations for the interface, experimental results with the hardware are presented. The experiments show that the interface is capable of operating mechanically with heavy loads of up to 40 kg. The proposed latch mechanism tolerates layers of dust of up to 2 mm. Thus, an electrical as well as mechanical connection in dusty environments is realized.

This paper presents the design of the robot AILA, a mobile dual-arm robot system developed as a research platform for investigating aspects of the currently booming multidisciplinary area of mobile manipulation. The robot integrates and allows in a single platform to perform research in most of the areas involved in autonomous robotics: navigation, mobile and dual-arm manipulation planning, active compliance and force control strategies, object recognition, scene representation, and semantic perception. AILA has 32 degrees of freedom, including 7-DOF arms, 4-DOF torso, 2-DOF head, and a mobile base equipped with six wheels, each of them with two degrees of freedom. The primary design goal was to achieve a lightweight arm construction with a payload-to-weight ratio greater than one. Besides, an adjustable body should sustain the dual-arm system providing an extended workspace. In addition, mobility is provided by means of a wheel-based mobile base. As a result, AILA's arms can lift 8kg and weigh 5.5kg, thus achieving a payload-to-weight ratio of 1.45. The paper will provide an overview of the design, especially in the mechatronics area, as well as of its realization, the sensors incorporated in the system, and its control software.

This paper presents the use of context-based predictions for the selection and on-line modification of the compliance of a robot manipulator. The work is partially inspired on current neuroscience hypotheses about the control of the human arm and the computational processes used by the brain. A first experiment uses inspiration from the classical neuroscience experiment of the Waiter Task. In the original experiment, the non-dominant human arm is holding a weight of 1 Kg. When this weight is unloaded by a self-generated action (with the dominant arm), it is observed that the non-dominant arm does not suffer perceptible postural changes. The reason arguably stems from the prediction of the forces occurring at the unloading, since the inherently delayed sensory feedback present in biological systems would not suffice to react in such a short notice as observed. The experiment is reproduced in a robotic platform by means of forward models and compliance adaption via stiffness control as speculated in neuroscience hypotheses. A second experiment uses context-based predictions to modify on-line the compliance of the robot manipulator. For this task, a Bayesian predictor in the form of a Relevance Vector Machine combines the use of prior knowledge and expected sensory feedback to correct for an erroneous compliance in the case of a falsely-predicted context. The results are combined in an architecture called Predictive Context-Based Adaptive Compliance (PCAC).

In this work, we describe a learning system that uses the CMA-ES method from evolutionary computation to optimize the morphology and the walking patterns for a complex legged robot simultaneously. Using simulation tools has the advantage that an optimization of robot morphology is possible before actually building the robot. Also, manually developing walking patterns for kinematically complex robots can be a challenging and time-consuming task. Both, the walking pattern and the morphology depend highly on each other to produce an energy-efficient and stable locomotion behaviour. In order to automate this design process, a learning system that generates, tests, and optimizes different walking patterns and morphologies is needed, as well as the ability to accurately simulate a robot and its environment. The evolutionary algorithm optimizes parameters that affect the trajectories of the robot's foot points, testing the resulting walking patterns in a physical simulation. The robot's limbs are controlled using inverse kinematics. In the future, the best solution evolved by this approach will be used for the mechanical construction of the real robot. Afterwards, the optimized walking patterns will be transferred to the real robot.

This paper describes the design and control methodology used for the development of a dual-arm manipulator as well as its deployment in a production scenario. Multi-modal and sensor-based manipulation strategies are used to guide the robot on its task to supervise and, when necessary, solve faulty situations in a production line. For that task the robot is equipped with two arms, aimed at providing the robot with total independence from the production line. In other words, no extra mechanical stoppers are mounted on the line to halt targeted objects, but the robot will employ both arms to (a) stop with one arm a carrier that holds an object to be inserted/replaced, and (b) use the second arm to handle such object. Besides, visual information from head and wrist-mounted cameras provide the robot with information such as the state of the production line, the unequivocal detection/recognition of the targeted objects, and the location of the target in order to guide the grasp.

Evolving a monolithic solution for complex robotic problems is hard. One of the reasons for this is the difficulty of defining a global fitness function that leads to a solution with desired operating properties. The problem with a global fitness function is that it may not reward intermediate solutions that would ultimately lead to the desired operating properties. A possible way to solve such a problem is to decompose the solution space into smaller subsolutions with lower number of intrinsic dimensions. In this paper, we apply the design principles of behavior based systems to decompose a complex robot control task into subsolutions and show how to incrementally modify the fitness function that (1) results in desired operating properties as the subsolutions are learned, and (2) avoids the need to learn the coordination of behaviors separately. We demonstrate our method by learning to control a quadrocopter flying vehicle.

An accurate motion model is an important component in modern-day robotic systems, but building such a model for a complex system often requires an appreciable amount of manual effort. In this paper we present a motion model representation, the dynamic Gaussian mixture model (DGMM), that alleviates the need to manually design the form of a motion model, and provides a direct means of incorporating auxiliary sensory data into the model. This representation and its accompanying algorithms are validated experimentally using an 8-legged kinematically complex robot, as well as a standard benchmark dataset. The presented method not only learns the robot's motion model, but also improves the model's accuracy by incorporating information about the terrain surrounding the robot.

In this paper we present a concept and an evaluation of an ape-like robot which is quite similar to its biological model. Aim of our project LittleApe is to build a small and extreme lightweight robot that is capable of walking on two and four legs as well as of changing from a four-legged posture to a two-legged posture, manipulating small objects, and which is also able to climb. LittleApe is modelled with attributes of a chimpanzee regarding limb proportions, spinal column, centre of mass, walking pattern, and range of motion. The concept of LittleApe is tested in simulation while building the real system. Two aspects were chosen to evaluate the concept described in detail within this paper. The first aspect comprises the use of an evolutionary method and the comparison of different morphologies. Based on the results from the first one, the second aspect deals with the manoeuvrability of the LittleApe robot.

Suspicion of water ice deposits in the lunar south-polar region have sparked new interest into the earth's smaller companion, and robotic crater sample return missions are being considered by a number of space agencies. The difficult terrain with an inclination of over 30°, eternal darkness and temperatures of less than -173°C make this a difficult task. In this paper we present a novel, bio-inspired light-weight system design, which demonstrates a possible approach for such a mission. The robot managed to come first in the Lunar Robotic Challenge (LRC), organised by the European Space Agency (ESA) in October 2008. Using a remote operated robot, we demonstrated to climb into and out of a lunar-like crater with inclination of more than 35° on loose substrate, and performed the collection and delivery of a 100 g soil sample without the aid of external illumination.

This paper describes the use of evolutionary algorithms to find an optimal solution for the parameters of an impedance controller represented as an artificial neural network (ANN). An impedance controller with force tracking capabilities has been evolved using evolutionary strategies which control the forces between a robotic manipulator and the environment. Simulation results show the controllerpsilas performance using a model of a two-link robot arm and a Hunt-Crossley non-linear model of the environment.

In recent years, neuroevolutionary methods have shown great promise in solving learning tasks, especially in domains that are stochastic, partially observable, and noisy. In this paper, we show how the Kalman filter can be exploited (1) to efficiently find an optimal solution (i. e. reducing the number of evaluations needed to find the solution), (2) to find solutions that are robust against noise, and (3) to recover or reconstruct missing state variables, traditionally known as state estimation in control engineering community. Our algorithm has been tested on the double pole balancing without velocities benchmark, and has achieved significantly better results on this benchmark than the published results of other algorithms to date.

For many complex Reinforcement Learning (RL) problems with large and continuous state spaces, neuroevolution has achieved promising results. This is especially true when there is noise in sensor and/or actuator signals. These results have mainly been obtained in offline learning settings, where the training and the evaluation phases of the systems are separated. In contrast, for online RL tasks, the actual performance of a system matters during its learning phase. In these tasks, neuroevolutionary systems are often impaired by their purely exploratory nature, meaning that they usually do not use (i.e. exploit) their knowledge of a single individual's performance to improve performance during learning. In this paper we describe modifications that significantly improve the online performance of the neuroevolutionary method Evolutionary Acquisition of Neural Topologies and discuss the results obtained in the Mountain Car benchmark.

In this paper we present a Common Genetic Encoding (CGE) for networks that can be applied to both direct and indirect encoding methods. As a direct encoding method, CGE allows the implicit evaluation of an encoded phenotype without the need to decode the phenotype from the genotype. On the other hand, one can easily decode the structure of a phenotype network, since its topology is implicitly encoded in the genotype's gene-order. Furthermore, we illustrate how CGE can be used for the indirect encoding of networks. CGE has useful properties that makes it suitable for evolving neural networks. A formal definition of the encoding is given, and some of the important properties of the encoding are proven such as its closure under mutation operators, its completeness in representing any phenotype network, and the existence of an algorithm that can evaluate any given phenotype without running into an infinite loop.

Several methods have been proposed for solving reinforcement learning (RL) problems. In addition to temporal difference (TD) methods, evolutionary algorithms (EA) are among the most promising approaches. The relative performance of these approaches in certain subdomains of the general RL problem remains an open question at this time. In addition to theoretical analysis, benchmarks are one of the most important tools for comparing different RL methods in certain problem domains. A recently proposed RL benchmark problem is the Keepaway benchmark, which is based on the RoboCup Soccer Simulator. This benchmark is one of the most challenging multiagent learning problems because its state-space is continuous and high dimensional, and both the sensors and actuators are noisy. In this paper we analyze the performance of the neuroevolutionary approach called evolutionary acquisition of neural topologies (EANT) in the Keepaway benchmark, and compare the results obtained using EANT with the results of other algorithms tested on the same benchmark.

This article describes: tests of fault tolerance of the eight-legged walking robot SCORMON in the event of leg loss. It evaluates different gaits, which are based on biological research on insect and arachnid walking and concludes with a discussion, what the best gait for the SCORPION system is, when 2 legs are lost It also includes a short introduction to the SCORPION robot and its biomimetic software approach and its performance in different terrain.

We build a snake-like robot for the inspection of areas that are difficult or dangerous to be accessed by human. Pictures from a camera at the robot's head are sent to a remote screen and can be monitored by a human operator. This human operator can also control the robot's motion, but not every single body part. Thus the operator only gives general directives and the robot will then able to follow autonomously the given directives. This is the semi-autonomous behavior of the robot. However, the robot must also be able to act fully autonomous when the contact to the operator is lost. After a short description of the robot we present a method which makes the operator control easy and also allows the robot to act fully autonomous. The autonomous motion control of the robot moving in a sewer pipe has been implemented.