Integration of task planning and motion control in a multi-robot assembly workcell

Abstract

Most of the existing multi-robot systems emphasize only one of two aspects, namely, plan generation and plan monitoring. There is a missing link between producing a multi-agent plan and its control during actual execution. The existence of this gap between planning and control precludes the possibility of automatically regenerating the execution plan to recover from unexpected error conditions and greatly limits the flexibility of the system. This paper describes a multi-robot task planning and execution system which successfully solves the integration problem by a two-phase five-layer structure. The concept of a finite-state automaton has been used as a tool to combine planning with simultaneous execution of parallel action. The developed system has been demonstrated for the assembly of a dishwasher power unit in a real experiment using a multi-robot assembly workcell consisting of an ADEPT I and a PUMA 560 robot placed close to each other so as to share a common workspace.