Planning Manipulation And Grasping Tasks With A Redundant Arm

This paper presents an approach to analysis and planning for autonomous manipulation tasks using a commercially available robotic arm and hand. We discuss hardware implementation, software architecture, and state machine behaviors. The inverse kinematics of the seven-degree-of-freedom manipulator are analytically derived with special attention to deterministically resolving the redundancy at the position level. Appropriate grasps are generated given the perceived position and orientation of the target object. Planning for the arm is accomplished in joint space using an RRT* implementation. Our implementation is able to detect an arbitrarily placed object in the workspace, generate a plan for the arm and hand avoiding obstacles, and execute the planned manipulation task. To show the efficacy of our approach, we present experiments in which the system autonomously grasps a canteen and a flashlight.