关键词: 昂贵多目标优化, Kriging代理模型, 多偏好物理规划, EGO算法, 自适应优化

Abstract: Aiming at the expensive multi-objective optimization problem with a large number of objectives and a high degree of dispersion of the Pareto fronts,most of the existing algorithms cannot use less function evaluation to obtain a high-quality Pareto front.Therefore,a surrogate-assisted multi-objective optimization algorithm based on multi-preference physical programming named Multi-preference Physical Programming Efficient Global Optimization (M3pEGO) was proposed.In this method,first sets up a preference matrix was built,and the multi-objective optimization problems was transformed into a single-objective optimization problem.Then the algorithm was combined with the Kriging surrogate model using Efficient Global Optimization (EGO) algorithm to achieve adaptive optimization.This method was compared with the ParEGO algorithm and multi-objective EGO algorithm through nine classical test functions.The results showed that the proposed algorithm had obvious advantages in solving expensive multi-objective optimization,especially problems with high dispersion of Pareto fronts.It could accurately fit to the true Pareto frontier after a finite number of iterations,and could obtain convergent and uniform non-dominated solution sets.

Key words: expensive multi-objective optimization, Kriging surrogate model, multi-preference physical programming, efficient global optimization algorithm, adaptive optimization