Job-shop scheduling with genetic programming
Kazuo Miyashita · 2000
In order to solve a real-time scheduling problem, a computationally intensive search-based optimization method is not practical, but the efficient dispatching rule that is well-customized for the specific problem at hand can be an effective problem solving method. A dispatching rule is scheduling heuristics that decide the sequence of operations to be executed at each resource in the scheduling problem. However, developing a customized dispatching rule for specific scheduling problems is an arduous task even for domain experts or researchers in the scheduling problem. In this research, the author views scheduling problems as multi-agent problem solving and proposes an approach for synthesizing the dispatching rule by means of Genetic Programming (GP). In the preliminary experiments, the author got the results showing that GP-based multi-agent dispatching scheduler outperformed the well-known dispatching rules.