RRT-based Motion Planning with Sampling in Redundancy Space for Robots with Anthropomorphic Arms

This work proposes a novel RRT-based motion planning method for robotic systems equipped with redundant anthropomorphic arms. A characteristic Redundancy space (ReRRT) for this type of robots is proposed for sampling at the position level to deal with the problem of Task-Constrained Motion Planning (TCMP) with other multiple constraints in realistic robotic manipulation applications. Based on the geometric feature of a redundant anthropomorphic arm with seven DOFs, the whole configuration space of a robot equipped with the anthropomorphic arm is decoupled to extract a redundancy space independent of the task space. The redundancy space is defined and spanned by the swivel angle of the arm and the variables which control the position and orientation of the arm base frame located at the shoulder. As a result, an explicitly intuitive sampling space, which has one-to-one mapping relationship with the task-constrained configuration space, can be found. One dedicated Inverse Kinematics (IK) algorithm is developed to directly map the samples back to the task-constrained configuration space without iterative modifications. Simulation studies were performed and comparisons with traditional Gradient Projection Method (GPM) and Rapidly exploring Random Tree (RRT) are drawn to illustrate that the proposed ReRRT is able to combine the advantages of the two methods. Finally, an experimental task of putting a cup on the table is performed on a humanoid robot COMAN to demonstrate the effectiveness of the proposed method in a more realistic task scenario.