Abstract

The present invention is embodied in a machine-implemented method of automatically correcting errors in an industrial schedule. The invention maintains a set of repair heuristics which may be applied to correct errors in the schedule. The system is initially trained by an expert who identifies problems in a set of trial schedules and applies the heuristics to correct the problems. The system records all steps, unsuccessful and successful, which were used by the expert to solve the problem. The recognized problem, the trial heuristics and the results of applying each heuristic are stored in a case library. When a sufficient number of trial schedules have been entered, the system may operate automatically. During automatic operation, the system automatically generates an initial schedule to solve a scheduling problem. Errors in the schedule are recognized either automatically or using input from an operator. Next, the case library is repeatedly searched for a solved case which most closely matches the present problem case and the previously successful heuristic is applied to solve the present problem until a solution is found. If no solution is found, the system takes the failed heuristics into account as a substitute for undefined contextual variables, which affect the schedule, in searching the case library.

Other References

• Ow et al.; Reactive Plan Revision; Proceedings of Seventh American Association of Artificial Intelligence; pp. 77-82;1988
• Yoshida et al.; A Dynamic Scheduling for Flexible Manufacturing Systems; Hierarchical Control and Dispatching by Heuristics; 1989; pp. 846-852; Proceedings of the 28th Conference on Decision and Contro