Intelligent scheduling agent for distributed decision-making

FMS is often described as 3 interconnected systems; machine set, material handling system, and storage and retrieval system. These subsystems can be modeled as a distributed decision-making system. We define a scheduling agent as an intelligent entity that can decide the allocation of production resources over time. It performs reactive and predictive scheduling, independently and concurrently, in real-time. We adopt two general forms of learning; individual learning, and group learning. Defined behavior can be achieved by one of these forms of learning, or by an interaction of the two. The agent consists of three principle components: predictive scheduling and reactive behavior, involved in individual learning, and reactive behavior towards other agents, involved in group learning. 3 neural nets are constructed for these learning. The connection weights are the memory units of a neural network; their values represent the current state of knowledge of the network. Learning in a neural network describes the way in which connecting weights are updated to generate a desired effect.