The IGD+ indicator based evolutionary algorithm for expensive multi-objective optimization problems

Due to the high experimental cost in function evaluation procedure, commonly used evolutionary algorithms are difficult to solve expensive multi-objective optimization problems. In this paper, we propose a new modified inverted generational distance based evolutionary algorithm to address this issue. The proposed method has introduced a Kriging model to replace the real model. Considering the better characteristics of the IGD
⁺
indicator, we select the performance indicator to balance the convergence and diversity of the population. In addition, we have embedded the mean value and mean squre error into the IGD
⁺
indicator. Meanwhile, two archives management strategy is introduced to store the non-dominated individuals and update the surrogate model. The proposed algorithm has been conducted on some benchmark test problems compared with four related algorithms. The experimental results have validated that the proposed algorithm is suitable for solving the expensive multi-objective optimization problems.