Manufacturing scheduling in decentralised holonic systems using artificial intelligence techniques

Reactive scheduling is used in decentralised systems, such as holonic or agent-based systems to obtain real-time feasible solutions for both assigning operations to processing machines and scheduling Material Handling (MH) resources. A holonic system using a decentralised approach for scheduling manufacturing tasks is considered in this study. Part of the holonic architecture, a global view component acts as an integrator for the individual decision-making processes and it is also used in the performance evaluation process of the holonic system. Artificial intelligence techniques are employed in the design of one optimal and three heuristic algorithms embedded in the evaluation module of the global view entity. Since there are no reported results of improvements made by learning mechanisms associated with MH holonic systems, this paper also investigates the addition of a Reinforcement Learning (RL) algorithm to the global view entity's evaluation module. Copyright © 2007 Inderscience Enterprises Ltd.