Anytime heuristic schedule repair in manufacturing industry

Although optimal schedules may be produced, the conditions under which those schedules remain optimal are unlikely to last for long, given the highly dynamic nature of the factory and the complexity of the informal, short-lived objectives and constraints which arise in practice, together with the shifting market conditions and the prevailing goal of pleasing the customer. The paper describes a software package which has been developed in collaboration with an industrial partner to permit schedule repair in response to dynamic conditions, both under the initiative of the human scheduler and automatically. A number of schedule heuristics are described to illustrate the problem of removing stockouts, together with some principles for designing repair heuristics. The performance of the heuristics is compared and a simple deliberation algorithm is presented for selecting dynamically between the heuristics depending on the rate of progress towards the goal. The results of experiments on simulated data are presented and discussed