Robust Multiobjective Optimization via State Transition Algorithm

In most real-world applications, uncertainties inevitably arise from inaccurate manufacturing, approximate models and varying environmental conditions. Similarly, uncertainties affect the multiobjective optimization process and directly influence the Pareto optimal front. In this paper, the uncertainties in decision variables are considered and a novel robust multiobjective optimization method is proposed to solve the problem. This method incorporates the mean effective concept to the multiobjective state transition algorithm based on the modified decomposition method(MOSTA/D). The mean effective concept calculating by the Monte Carlo method is used to measure the robustness of solutions and MOSTA/D is used to solve Multiobjective optimization problems. Experimental results on a group of benchmark test functions demonstrate the superiority of the proposed method in terms of both solutions' optimality and robustness under disturbances.