Planning multi-robot grasping motions

In this work, a motion planning approach for multi-robot grasping problems is presented. The proposed Multi-Robot-RRT planner controls multiple IK-RRT instances and evaluates grasp qualities for resulting multi-handed grasping con gurations until a global solution is found. Furthermore the Multi-Robot-RRT concept avoids deadlock situations which may appear due to the decomposed planning approach. RRT-like search in phase space is applied to the results to generate collision-free phases for each motion that allow a simultaneous execution of planned trajectories. Two experiments are presented to demonstrate the usability of the proposed algorithms for planning problems covering up to 60 DoF of three humanoid robots. The planners enable humanoid robots to cooperatively grasp large objects by combining the search for solutions to the three main problems of multi-robot grasping to one integrated planning concept: building feasible bimanual grasping con gurations for each robot, searching a stable multi-robot grasping setup and planning collision-free motions.