Implementing a multi-agent systems approach to collaborative autonomous manufacturing operations

Remote space operations is a very demanding domain for software systems, requiring the characteristics of high availability, robustness, autonomy, real-time or near real-time response times, ease of reprogramming, and low computational footprint. A further desirable characteristic is the ability for remote systems or vehicles to collaborate to achieve tasks such as exploration or remote processing or sensing. Intelligent agent technology has advanced to the point where it is now finding application in robotic manufacturing, such as the application described in this paper. The authors believe that many of the lessons being learned in the collaborative manufacturing domain have relevance to space operations. This paper describes a novel agent-based execution architecture. Integral to this architecture is the separation of part processing concerns from manufacturing concerns. Thus we have part agents which are able to issue and track resource-independent part-processing requests. An interface agent then transforms a resource-independent request into one or more resource-dependent requests and dispatches them to the appropriate manufacturing agents. Execution of the dispatched requests then proceeds autonomously, taking into account safety constraints. We also describe an implementation of this architecture for a robotic assembly cell located at the University of Cambridge using JACK Intelligent Agents/sup TM/. The paper concludes with the lessons learned from this experiment, and highlights those of relevance to the domain of space operations.