Constrained multi-objective wolf pack algorithm based on cooperative optimization and tiered hunting

Abstract

In tackling constrained multi-objective optimization problems, conventional multi-objective wolf pack algorithm often exhibits poor feasibility and diversity, primarily due to the lack of explicit constraint-handling mechanisms and strong reliance on the leader wolf leading to population aggregation. To overcome these issues, this paper proposes a constrained multi-objective wolf pack algorithm based on cooperative optimization and tiered hunting (CTCMOWPA). The cooperative optimization technique is designed to balance exploration of the infeasible region and exploitation within the feasible region. It maintains two cooperative populations with distinct roles. The main population preserves feasibility by applying the CDP. The secondary population first explores the unconstrained Pareto Front and then employs an adaptive


$$\varepsilon$$

ε



constraint-handling method to guide the search toward feasible regions. Moreover, information exchange between these populations enhances adaptability to diverse constraints. Additionally, a tiered hunting strategy categorizes wolves into elite wolves characterized by lower fitness that besiege the leader wolf to inherit its information, and exploration wolves distinguished by higher fitness that are dually guided by the leader and elite wolves, thereby suppressing aggregation while maintaining diversity and convergence. Experimental comparisons with six advanced algorithms on three benchmark suites demonstrate that CTCMOWPA achieves the best IGD values on 19 out of 38 test problems and outperforms the comparison algorithms in feasibility and diversity.