Base Placement and Human-like Motion Planning for a Humanoid Mobile Dual-arm Manipulator System

At present, robots' autonomous ability is still very weak, hence improving the operation ability and human-robot interaction & collaboration capabilities is one of the primary goals for developing the anthropomorphic robots. In order to improve the human-like autonomy of anthropomorphic robots in dynamic and unstructured environments, namely, predictable and human-like motion planning, this paper proposes an on-line motion planning method for a humanoid mobile dual-arm manipulator system. It involves two aspects: autonomous placement of the mobile base (MB) and human-like motion planning of the MB and two arms. In particular, the two aspects are achieved by introducing two relative inverse dynamic reachability maps: RiDRM and LiDRM. Since the construction of RiDRM and LiDRM is mainly done offline, the online efficiency is guaranteed. In fact, if an obstacle is detected online, the maps RiDRM and LiDRM are updated by only disabling the candidate end-poses which are in collision. Furthermore, predictable and human-like movements are achieved by optimizing a merit function that considers five criteria simultaneously. The simulation results verify the effectiveness of the proposed method.