Agent-Based Adaptive Production Scheduling — A Study on Cooperative-Competition in Federated Agent Architecture
An increasingly popular method of improving the performance of complex systems operating in dynamic environments involves modeling such systems as social networks made up of a community of agents working together based on some basic principles of social interaction. However, this paradigm is not without its challenges brought about by the need for autonomy of agents in the system. While some problems can be solved by making the interaction protocol either strictly competitive or strictly cooperative, some other models require the system to incorporate both interaction schemes for improved performance. In this paper, we study how the seemingly contradictory effects of these two behaviours can be exploited for distributed problem solving by considering a flexible job shop scheduling problem in a dynamic order environment. The system is modeled using federated agent architecture. We implement a simple auction mechanism at each processing center and a global reinforcement learning mechanism to minimize cost contents in the system. Results of simulations using the cooperative-competition approach and the strictly competitive model are presented. Simulation results show that there were improvements in cost objectives of the system when the various processing centers cooperated through the learning mechanism, which also provides for adaptation of the system to a stream of random orders.
KeywordsSchedule Problem Multiagent System Inventory Cost Task Sequence Competitive Strategy
Unable to display preview. Download preview PDF.
- Aydin, M.E. and Öztemel, E.: Dynamic job-shop scheduling using reinforcement learning agents, Robotics and Autonomous Systems, Volume 33, Issues 2–3, 30 November 2000, Pages 169–178Google Scholar
- Baguchi, T.P.: Multiobjective Scheduling by Genetic Algorithms, Kluwer Academic Publishers, USA, 1999Google Scholar
- Pinedo, M.: Scheduling, Theory Algorithms and Systems (2nd Edition). Prentice Hall, USA. 2002Google Scholar
- Sandholm, T.: Distributed Rational Decision Making. Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence (Ed. By Weiss, G.). The MIT Press, USA. 1999.201–258.Google Scholar
- Sandholm, T.: An Algorithm for Optimal Winner Determination. in Combinatorial Auctions. IJCAI 1999, 542–547.Google Scholar
- Shen, W., Nome, D.H., Barthes, J.A.: Multi-agent system for concurrent intelligent design and manufacturing. Taylor & Francis Inc. USA. 2001Google Scholar
- Walsh, W.E., Wellman, M.P.: A Market Protocol for Decentralized Task Allocation. Proceedings of the Third International Conference on Multi-Agent Systems, IEEE Computer Society 1998. 325–332Google Scholar
- Weiss, G. (Ed.): A Modern Approach to Distributed Artificial Intelligence, The MIT Press, USA. 1999Google Scholar