Using Multi-agent System for Dynamic Job Shop Scheduling1

Today’s industries need more flexible scheduling systems able to produce new valid schedule in response to the modifications concerning orders, production processes and deliveries of materials. This paper introduces a multi-agent system applied to a job shop dynamic scheduling problem in which new production orders or deliveries arrive continuously and affect the already scheduled plan. We have solved the problem by: i) coupling reactive and pro-active agent behavior; and ii) implementing a stochastic method simulated annealing into agent’s behavior. The job shop scheduling system is implemented using various types of agents whose interactions make the global state of the system move from a solution to another by continuously adapting to the changes from the environment. In this perspective, the interactions between the agents representing the client job orders, the production centers and the material stocks result in the assignment of operations and the plan for stock movements. Our experimental results show that, by modifying the classical agent-based message scheme, the integration of stochastic approach and multi-agent based technology could improve dynamic scheduling problems for a small to medium size problem space. 1 ICEIS 2002, Fourth International Conference on Enterprise Information Systems, Ciudad Real, Spain, April 2002