Towards Real-Time Execution of Motion Tasks

The goal of the research described in this paper is to build robotic systems that can execute motion tasks in real time. We present two ideas towards this goal: a real-time path planner for static three-dimensional configuration spaces and a new framework for the integration of planning and control. The planner is based on a new class of cells, slippery cells, incorporated in an approximate cell decomposition algorithm. The elastic band concept is proposed to provide an effective link between planning and execution. With elastic bands, a path is treated as a flexible entity. The initial configuration of the elastic band is the path provided by the planner. The shape of the elastic dynamically evolves during execution, using sensory data about the environment.