Solving Generic Scheduling Problems with a Distributed Genetic Algorithm

This paper describes a Distributed Genetic Algorithm (DGA) which has been used to solve generic scheduling problems. The GA based scheduler allows the user to deene and solve any scheduling problem. It does this using a Scheduling Description Language (SDL). The sort of problem that it might tackle are: job-shop scheduling (JJS), time-tabling, resource sequencing etc. We describe a unique chromosome coding scheme that allows simple representation, straightforward chromosome recombination and fast schedule building and therefore evaluation times. A comparative study has been made of the DGA, random search and a heuristic method of scheduling using 100 very large scale problems; problems of the order of 500 tasks. This is the rst study of its kind to look at problems of this scale. It was found that, although it is possible to reduce the makespan of a schedule by about 2 5 of a randomly generated solution using dispatching rules, only the DGA produced solutions that had as high as a 3 5 reduction.