General heuristic procedures and solution strategies for FMS scheduling

Abstract

This paper presents two new heuristic procedures for FMS scheduling. The heuristics decompose the very complex scheduling problem into a series of relatively easily handled subproblems, and solve them using MILP models and heuristics. Unlike traditional 'routing then sequencing' methods, both procedures consider constraints not only on machines but also on other critical resources, ensuring practical feasibility of the resulting schedules while reducing complexity in the solution process. The first heuristic, SEDEC, adopts an improved 'routing then sequencing' method with a further step for allocating other resources. The second procedure, CODEC, represents a new scheduling strategy which emphasizes interconnections between subproblems in addition to the solution of the individual subproblems themselves. Computational experiments are carried out in various scheduling environments to compare the performance of the two heuristic procedures with a large MILP model which considers the whole FMS scheduling problem. The results show that the heuristic procedures can generate schedules with optimality close to the solution of the MILP method, in a much shorter time. Comparing the two heuristic approaches, CODEC, representing a new solution strategy, performs significantly better on average than SEDEC, which is based on the traditional strategy. This is observed in all the FMS scheduling environments tested, especially when the problem is large, complex or with tight resource constraints. Analysis of the results also suggests that CODEC may be improved further by investigating the way the subproblems are solved and the way the routing subproblem is reformulated. The CODEC solution strategy provides a general framework for scheduling in a wide range of FMS environments.