Simplex model based evolutionary algorithm for dynamic multi-objective optimization

Abstract

Most real-world problems involve objectives, constraints and parameters which constantly change with time. Treating such problems as static problems requires knowledge of the prior time but the computational cost is still high. In this paper, a simplex model based evolutionary algorithm is proposed for dynamic multi-objective optimization, which uses a modified simplex model to predict the optimal solutions (in variable space) of the next time step. Thereafter, a modified evolutionary algorithm which borrows ideas from particle swarm optimization is applied to solve multi-objective problems when the time step is fixed. This method is tested and compared on a set of benchmarks. The results show that the method can effectively track varying Pareto fronts over time. © 2011 Springer-Verlag.