Solving multi-objective job shop problem using nature-based algorithms: new Pareto approximation features

Abstract

In this paper the job shop scheduling problem (JSP) with minimizing two criteria simultaneously is considered. JSP is frequently used model in real world applications of combinatorial optimization. Multi-objective job shop problems (MOJSP) were rarely studied. We implement and compare two multi-agent nature-based methods, namely ant colony optimization (ACO) and genetic algorithm (GA) for MOJSP. Both of those methods employ certain technique, taken from the multi-criteria decision analysis in order to establish ranking of solutions. ACO and GA differ in a method of keeping information about previously found solutions and their quality, which affects the course of the search. In result, new features of Pareto approximations provided by said algorithms are observed: aside from the slight superiority of the ACO method the Pareto frontier approximations provided by both methods are disjoint sets. Thus, both methods can be used to search mutually exclusive areas of the Pareto frontier.