Motion Planning for Humanoid Robots

Motion Planning for Humanoid Robots

A humanoid robot operating in a human-centered environment needs a component to plan collision-free motions in a fast and robust manner. Based on randomized algorithms, such as Rapidly Exploring Random Trees (RRT), we are developing algorithms for collision-free motion planning for single and dual grasping, re-grasping and dual-arm manipulation tasks. In addition, we are investigating grasping pipelines combining several tasks, such as finding a feasible grasp, solving the inverse kinematics and searching the configuration space for collision-free trajectories. The planning algorithms are evaluated and integrated on the humanoid robots of the ARMAR series which are developed by the High Performance Humanoid Technologies lab (H²T), KIT, Germany.

Most of the developed algorithms become part of the open source C++ software framework Simox, which can be downloaded from gitlab. The simulation and motion planning toolbox offers the possibility to simulate complex robot systems (VirtualRobot), to realize grasp planning algorithms (GraspStudio) and to plan collision-free motions (Saba).

Videos