High performance computing for dynamic multi-objective optimisation

摘要

In this paper a generic parallel procedure for dynamic problems using evolutionary algorithms is presented. In dynamic multi-objective problems, the objective functions, the constraints and hence, also the solutions, can change over time and usually demand to be solved online. Thus, high performance computing approaches, such as parallel processing, should be applied to these problems to meet the solution constraints and quality requirements. Taking this into account, we introduce a generic parallel procedure for multi-objective evolutionary algorithms, through a master-slave paradigm. This generic parallel procedure is used to compare the parallel processing of a few multi-objective optimisation evolutionary algorithms: our proposed algorithms, SFGA and SFGA2, in conjunction with SPEA2 and NSGA-II. We also give a model to understand the benefits of parallel processing in dynamic multi-objective problems and the speedup results observed in our experiments.