Collision-free walking in cluttered environments is still an open issue for humanoids. Most current approaches use heuristics with large safety margins to plan the robot's motion. That way, the chance of collisions can be greatly reduced but the robot movements are artificially limited. In this context, we extend our framework for motion generation and whole-body collision-avoidance by an on-line predictive kinematic parameter evaluation and optimization: we propose to evaluate the parameter set describing the walking pattern by integrating the full kinematic model of the robot. Initial parameter sets, which are kinematically infeasible due to kinematic limits or collisions, can be identified and adapted before the motion is executed. Starting with a feasible solution, the parameter set is optimized using a gradient method. Since the method is applied before each step, while the robot is executing the previous step, it is very reactive to changes in the environment or in the user input. The optimization method is not limited to a specific walking pattern representation, but it is applicable to different representations. We want to emphasize its suitability for real-time control. The optimization can be stopped if it exceeds a predetermined time budget. In that case, an executable but suboptimal result is used. The method is presented with simulation results obtained with our multi-body simulation. We have also validated the real-time performance in experiments with our humanoid Lola.

Selective tasks such as harvesting or spraying of single crops are a promising research topic in agricultural automation. Inspired by industrial production, an obvious approach is to use robot manipulators in greenhouses and orchards. To exploit the potential of redundant manipulators in particular, advanced motion planning algorithms are needed. While harvesting, a new trajectory for every fruit has to be planned. Although the scenario is similar for every fruit, it is unique for each harvesting sequence. In this paper we present an efficient online planning approach which takes advantage of a simplified environment model. However, the generated trajectory is not optimal in general w.r.t. joint velocities or might even be unfeasible. Thus, we introduce an optional global offline optimization scheme which is able to find optimal trajectories in a few seconds and takes advantage of the heuristic planning as initial guess. We apply the proposed scheme to a 9-DOF agricultural manipulator for sweet-pepper harvesting and evaluate our method by extensive tests with fruit positions based on real measurements.

In this paper we present a step-planner embedded in a framework which enables a humanoid robot to navigate among obstacles, exploiting its overall capacities. The system allows the robot to react to changes of user input or changes in a dynamic environment in real-time while walking at reasonable speeds. The proposed method relies neither on external sensors nor on color coding or textured surfaces. The key idea is to use a fast collision model based on swept-sphere-volumes (SSVs) for real-time generation of collision-free footsteps and whole-body trajectories. Using a SSV-based 3D approximation in all control modules enables the robot to avoid collisions with itself and the environment. Obstacles are detected with an on-board RGB-D sensor while the robot navigates through an environment which is not known in advance. A step-planner reacts to high-level user commands like desired velocity and direction or changes in the environment within less than a step. Instead of investigating only the footholds an articulated 3D approximation of the lower leg and the foot is considered to find feasible and optimal footstep locations. Additionally, it provides an initial solution for the swing-foot movement. Finally, collision-free swing-foot trajectories are created in real-time in the feedback control layer using all the foot's degrees of freedom. We validated this approach in experiments with our robot Lola.

This paper introduces a new state estimator for biped robots fusing encoder, inertial and force torque measurements. The estimator is implemented as a Kalman filter that uses the dynamical model of the linear inverted pendulum with the center of mass (CoM) state as output. In order to compensate for disturbances and model errors we extend the model by a state for the external force and an additional input which is calculated from the dynamics error in pattern generation. Several simulation results underline the effectiveness of the proposed filter and show its robustness against disturbances. Experimental results and an application example validate the method under real world conditions.

The drive mechanism of many robot joints are composed of an electrical actuator and a gear transmission. Besides actuator dynamics and gear elasticity, friction effects are of particular importance for accurate dynamic modeling. This paper presents the design and development of a modular testbed for experimental friction identification in modular robot drives. We have used this testbed to investigate modules that were developed at our institute for the humanoid robot Lola and an agricultural manipulator. We discovered that a friction law, similar to a law proposed in the literature, can be very accurately fitted to our measurements.

Determining the optimal solution for the inverse kinematics of redundant robots has been the focus of much previous research. Instantaneous approaches are computationally efficient, but may cause high joint velocities due to their local character. In this paper, we present an efficient implementation of a global approach following Pontryagin's Minimum Principle for online calculation. Within a moving horizon, we exploit the decoupled structure of the resulting optimal control problem by using the conjugate gradient method for solving the nonlinear dynamic problem. Different examples of cost functionals are presented and the real-time capability is shown by applying this approach to a 9-DOF redundant manipulator.

The ability to avoid collisions is crucial for locomotion in cluttered environments. It is not enough to plan collision-free movements in advance when the environment is dynamic and not precisely known. We developed a new method which generates locally optimized trajectories online during the feedback control in order to dynamically avoid obstacles. This method successfully combines a local potential field method with a heuristic based on height and width of an obstacle to avoid collisions. The program's main feature is the integration of obstacles into the framework designed for self-collision avoidance presented in [1] and the collisions avoidance in task-space. We show experimental results validating the method.

Pushing, soft ground contact and other unknown large disturbances can cause humanoid robots to fall. In order to face such problems, an accurate and fast model of the robot is necessary to estimate the state at a certain time in the future using only current measurements. We approximate the robot by a three-mass model with two degrees of freedom. Unilateral compliant contacts and a closed-loop control similar to the one used on the real robot are included additionally. Utilizing this model it is possible to predict the robot's behavior and calculate compensating motions for different disturbance cases. Furthermore, it is possible to integrate the estimation results into the real-time control system of our biped LOLA. Simulation and experimental results demonstrate the effectiveness and the advantages of the proposed method.

In this paper we describe an event-based walking control system for biped robots. Classically, the control of biped robots has been separated into walking pattern generation and stabilizing control. While this is an effective strategy in well-known environments, walking in rough, unmodelled terrain can easily destabilize the system. Findings from neurobiology suggest that gait generation in animals and humans is strongly linked to sensory signals indicating certain events in the gait cycle. We therefore propose an event-based controller that triggers phase transitions based on the sensed walking state instead of only relying on time-based reference trajectories. We have implemented the proposed approach on our humanoid robot Lola. We present simulations and experimental results demonstrating the effectiveness of the approach when walking over unknown obstacles.

To simplify the development of new trajectories for a humanoid robot with many joints, the ability to check for potential collisions is crucial. More importantly, to enable on line generation of general motions, such as whole-body-motion, collision avoidance must be employed, which is usually more challenging. In this paper, a model-based and efficient on-line distance calculation between links of the robot is proposed. To avoid collisions the results are projected into the null space term of the inverse-kinematics which is based on the redundancy resolution framework of LIEGEOIS. Additionally, a novel method to reduce the vertical angular momentum for a walking biped robot which uses the arm motion is proposed. It reduces the vertical angular momentum of the walking robot especially for larger steps. Since the method also works in the null-space, it can easily be combined with the proposed collision avoidance scheme. Finally, results for the proposed methods from simulations and experiments with the robot Lola are shown.

This paper presents our new 25-DoF humanoid walking robot LOLA. The goal of our research is to realize a fast, human-like walking motion (target speed: 5 km/h). Furthermore, we want to increase the robot's autonomous, vision-guided walking capabilities. The robot has 25 degrees of freedom, including 7-DoF legs with actively driven toe joints. It is characterized by its lightweight construction, a modular, multi-sensory joint design with brushless motors and an electronics architecture using decentralized joint controllers. Special emphasis was put on an improved mass distribution of the leg apparatus to achieve good dynamic performance. The sensor system comprises absolute angular sensors in all links, two custom-made force/torque sensors in the feet and a high-precision inertial sensor in the upper body. Trajectory generation and control system aim at faster, more flexible, and more robust walking patterns.

This paper describes a real-time walking control system developed for the biped robots JOHNNIE and LOLA. Walking trajectories are planned on-line using a simplified robot model and modified by a stabilizing controller. The controller uses hybrid position/force control in task space based on a resolved motion rate scheme. Inertial stabilization is achieved by modifying the contact force trajectories. The paper includes an analysis of the dynamics of controlled bipeds, which is the basis for the proposed control system. The system was tested both in forward dynamics simulations and in experiments with JOHNNIE.

Dynamics simulation is a valuable tool for biped robot hardware and controller design. Due to the complexity of biped walking robots, simulation times can be quite high. It is therefore important to identify the key components in modeling the system so that the phenomena of interest can be correctly predicted at a reasonable numerical cost. In this paper a detailed and a reduced simulation of the biped robot JOHNNIE are presented and results are compared with experimental data. The comparison shows that results for both models correctly predict global dynamics of the system, while some phenomena are only captured by the more complex model

The paper presents the performance enhanced humanoid robot LOLA which is currently being manufactured. The goal of the project is the realization of a fast, human-like walking motion. The robot is characterized by its lightweight construction, a modular, multi-sensory joint design with brushless motors and an electronics architecture using decentralized joint controllers. The fusion of motor, gear and sensors into a highly integrated, mechatronic joint module has several advantages for the whole system, including high power density, good dynamic performance and reliability. Additional degrees of freedom are introduced in elbow, waist and toes. Linear actuators are employed for the knee joints for a better mass distribution in the legs