A local correlation estimation surrogate-assisted bi-objective evolutionary algorithm for heterogeneous objectives

Most existing evolutionary algorithms for solving bi-objective optimization problems are based on the assumption that both objectives can be evaluated once at the same time. However, in real-world optimization problems, there is a significant difference in function evaluation time between objectives, and these problems are defined as the problems with heterogeneous objectives. To utilize the latency of objectives, we propose a local correlation estimation based surrogate-assisted bi-objective evolutionary algorithm for problems with heterogeneous objectives. In the proposed algorithm, surrogate models are employed to approximate the objective functions. The proposed local correlation estimation (LCE) method is used to analyze the correlation between objectives within a local region, guiding the search direction for one objective while identifying promising solutions for the other objective. Finally, the ablation experiment for the LCE method validates the effectiveness of the proposed strategy, and the comparative results on various expensive bi-objective test problems demonstrate that the proposed algorithm is promising on heterogeneous bi-objective optimization problems.