Real-time Adaptive Mobile Manipulator Motion Planning

Abstract

This video demonstrates a real-time adaptive motion planner for a mobile manipulator to accomplish place-to-place tasks in a dynamic environment with obstacles of unknown motion. Paths and trajectories are planned simultaneously as the robot moves and globally subject to some optimization criteria based on evolutionary computation. The robot always follows the current best trajectory with respect to predictions of obstacle motions through sensing. At any time the robot may switch seamlessly to a better trajectory as the planner continues to improve or adapt trajectories to the changing environment. To minimize energy and time, the planned arm and base trajectories are loosely-coupled so that the arm may stop its motion (relative to base) for some period while the base moves, or vice versa, in handling obstacles.