Generating human-like movements for robotic arms

• Two new criteria are constructed to resolve the kinematic redundancy of the human arm.
• A new optimal sampling based motion planner is proposed.
• Methods to generate human-like reaching and grasping movements are proposed.
• Both the end-effector and arm postures of the generated movements are human-like.

In this paper, the generation of trajectories of both end-effector and joints for human-like reaching and grasping motions is studied. In reaching movement, the human-like end-effector trajectory is obtained based on the minimum jerk model. A total potential energy criterion is constructed to resolve the kinematic redundancy of human arm in the target position. Gradient Projection Method (GPM) is adopted to trace the human-like end-effector trajectory while minimizing the total potential energy to generate the human-like joint trajectory. In grasping movement, the total potential energy and wrist discomfort are synthesized to resolve the kinematic redundancy of human arm in the target position and orientation. A new planner, Gradient Projection Method based Rapidly-exploring Random Tree (GPM-RRT) algorithm, is proposed to generate the human-like end-effector trajectory by goal biasing and the human-like joint trajectory by minimizing the synthesis of the total potential energy and wrist discomfort. The criteria and algorithm are verified by simulations and experiments.