Robust Multi-objective Optimization based on the idea of multi-tasking and knowledge transfer

Abstract

In practical application, in order to make the solution obtained from optimization problems have higher practical significance and value in the real world, it is necessary to develop effective optimization methods under uncertainty. The multi-objective optimization method based on evolutionary algorithm is devoted to improving the performance of multi-objective optimization and dealing with uncertainty disturbance, which is very suitable for the related problems to be studied in this paper. In this paper, a new multi-objective evolutionary optimization method is proposed for multi-objective optimization problems with uncertainty in decision space, which aims to find a set of robust optimal solutions and achieve the optimal balance between optimality and robustness. The proposed algorithm uses the idea of multi-tasking and knowledge transfer and uses one population to solve two optimization problems. In this paper, a vector archive updating strategy based on memory mechanism is proposed to maintain the robustness and optimality of Pareto solution set, measure the average and worst case of solutions under different disturbance degrees, and evaluate each solution comprehensively. Experiments show that the proposed robust multi-objective optimization method can obtain a set of robust Pareto optimal solutions in the presence of disturbances, and its robust performance is better than other algorithms in most test problems.