Dynamic grid-based uniform search for solving constrained multiobjective optimization problems

When solving constrained multiobjective optimization problems (CMOPs), it is important to uniformly explore the promising regions that are not dominated by feasible solutions, and this can effectively avoid the loss of the Pareto front fragments. To achieve this, we propose a grid-based uniform search (GUS) to guide the current population to search the promising areas uniformly in this paper. Therein, the promising areas are divided into a number of grids, which are then fully explored by the individuals located in them. In the process of reducing the population size, the individuals with the largest constraint violations in the most crowded grids are removed one by one. To balance the local search and the global search, we dynamically reduce the number of divided grids in GUS with the increase of evolutionary iterations. Embedding the dynamic GUS in evolutionary algorithm, we design a new constrained algorithms for CMOPs. Experimental results show that the proposed algorithm performs better than other state-of-the-art constrained evolutionary multiobjective optimization algorithms in dealing with different CMOPs.

Automated summary

When solving constrained multiobjective optimization problems, a grid-based uniform search method is proposed to balance local search and global search by dynamically adjusting the number of grids, achieving better results than other algorithms.