This paper presents a surrogate optimization method for the expensive black-box multi-objective optimization problems,which balances local and global searches using multiple sampling strategies in iterations.Under the framework of the surrogate optimization,this method employs a radial basis function interpolation model to approximate.During the iterations of the algorithm,a local search assisted target value strategy is implemented,in which multiple approximate Pareto fronts are constructed,multiple vectors filling the gaps of the approximate Pareto fronts are selected as the candidates of the target value,a new target value is determined based on the evaluated decision vectors and the existing target values,and a local search is undertaken around the corresponding candidate of the sample point.Moreover,a clustering method is used in the surrogate optimization sampling to enhance the diversity of the sample points.Numerical experiments on 58 standard test problems consisting of both low-and high-dimensional ones,as well as two practical problems,demonstrate the effectiveness of the proposed algorithm.
维普
万方数据