A Motion Planning Based To appear in the Proceedings of 1997 ICRAApproach for Inverse Kinematics

We propose a new approach to solving the point-to-point inverse kinematics problem for highly redundant manipulators. It is inspired by recent motion planning research and explicitly takes into account constraints due to joint limits and self-collisions. Central to our approach is the novel notion of kinematic roadmap for a manipulator. The kinematic roadmap captures the connectivity of the connguration space of a ma-nipulator in a nite graph like structure. The standard formulationof inverse kinematics problem is then solved using this roadmap. Our current implementation , based on Ariadne's Clew Algorithm BA + 93], is composed of two sub-algorithms: EXPLORE, an appealingly simple algorithm that builds the kinematic roadmap by placing landmarks in the connguration space; and SEARCH, a local planner, that uses this roadmap to reach the desired end-eeector conngura-tion. Our implementation of SEARCH is an extremely eecient closed form solution, albeit local, to inverse kinematics that exploits the serial kinematic structure of serial manipulator arms. Initial experiments with a 7-dof manipulator have been extremely successful.