Orientation-based reachability map for robot base placement

Mobile humanoid robots have the capability of accomplishing complex tasks in human dominated environments. In order to execute a task with its arm, a robot needs to place its base reasonably first. In previous work, a positionbased reachability map solved this problem by matching a precomputed inverse kinematics database with discretized versions of the task path. However, its capability is restricted to the end effector with which the database was computed. This paper proposes an orientation-based reachability map which supports on-line end effector frame extensions. The added extension capability enables robots to handle partially constrained task paths with different tool frames, which would be non-trivial for previous methods without recomputing a reachability map. We discuss the differences between the new method and its predecessors, and analyze extension transformation matrix to provide mathematical proof for the extension capability of our reachability map. Our experimental results show that the orientation-based reachability map is as fast as the positionbased reachability map while benefiting from the additional extension capability.