A Novel Dynamic Multi-objective Robust Evolutionary Optimization Method

Traditional methods solving dynamic multi-objective optimization problems (DMOPs) often trigger the evolution process again to find the Pareto-optimal solutions as soon as new environment appears. This may lead to larger computation and resources costs, even unable to perform the optimum solution in the limited time. Therefore, a novel evolutionary optimization method is proposed looking for dynamic robust Pareto-optimal solution sets, which are the Pareto-optimal solutions for certain environment. They can approximate to the true Pareto fronts in following consecutive dynamic environments along a certain satisfaction threshold, and directly be used as Pareto solutions of these environments so as to reduce the computation cost. Two metrics including time robustness and performance robustness are presented to measure the environmental adaptability of Pareto-optimal solutions. Subsequently, they are transformed into two kinds of robust optimization models. Multi-objective evolutionary algorithm based on decomposition and penalty-parameter less constraint handling method are introduced to form the decomposition-based dynamic multi-objective robust evolutionary optimization method. Especially, a moving average prediction model is adopted to realize multi-dimensional time series prediction of these solutions. In term of eight benchmark functions and two novel metrics, the simulation results indicate that the proposed method can obtain the robust Pareto-optimal solutions meeting the need of decision makers with more average survive time. Copyright © 2017 Acta Automatica Sinica. All rights reserved.