The next generation of intelligent and flexible manufacturing systems must be effective and efficient in manufacturing products in small batches, even down to lot sizes of one unit. In order to achieve this object, it is necessary to move factory control from the more rigid centralizeed and hierarchical manufacturing computer control architectures (MCCA) to fully decentralized, heterarchical architectures which can reconfigure dynamically depending on production requirements. In addition to taking advantage of the decentralized nature of the manufacturing process, a truly decentralized control system is more extensible and fault tolerant than a centralized or a hierarchical one. However, this comes with the price of designing the rules of interaction and cooperation among the agents comprising such a control systemt that ensure a satisfactory performance of the underlying manufacturing system.

We present the preliminary results of our investigation into heterarchical scheduling for flexible manufacturing systems. In our work, the workcells comprising the manufacturing process and the products to be generated are modeled as autonomous agents which interact dynamically to generate the production schedule for each product unit. The interaction scheme combines production reservation with a bidding mechanism using the contract net protocal. The effectiveness of this approach is demonstrated using simulation experiments that compare its performance to heuristic dispatching rules commonly used in industry.

Full Paper (zipped postscript)

Last updated: 5 February, 1997 by Anthony Alford alford@vuse.vanderbilt.edu