Neuro-Genetic Order Acceptance in a Job Shop setting

In this paper a new neuro-genetic architecture is presented that solves a profit oriented dynamic job shop problem. In the job shop order acceptance and scheduling problem new jobs arrive continuously and because of insufficient job shop capacity, a selection has to be made among the offered jobs. The goal is to find an order acceptance policy, which is supported by a scheduling policy, that maximizes the long-term profit for the job shop. The acceptance policy is learned through training a neural network using reinforcement learning and the scheduling policy is based on a genetic search driven by the same neural network. The obtained acceptance and scheduling policy is found to outperform two heuristic policies under various manufacturing en-