The solution to constrained multi-objective optimization problems (CMOPs) requires optimizing the objective functions while satisfying the constraint conditions. To effectively address CMOPs, algorithms must balance objectives and constraints. However, the limited adaptability of specific constraint-handling techniques (CHTs) has hindered the widespread applicability of constrained multi-objective evolutionary algorithms (CMOEAs). To overcome this limitation, this article proposes a population state-based CMOEA. First, a model is developed to identify population states based on the positions of the primary and auxiliary populations. Tailored environmental selection models are then designed for the auxiliary population according to different states, enabling them to guide the evolution of the main population more effectively. By dynamizing the CHTs, the proposed algorithm can adapt to a broader and more complex range of CMOPs. Additionally, state-specific optimal individual selection methods are introduced, allowing the auxiliary population to escape local optima and accelerate exploration. A simple yet effective resource allocation model is incorporated to address the potential computational resource waste associated with dual populations, enhancing the resource utilization. Comprehensive tests, including comparisons with seven state-of-the-art algorithms, were conducted on 47 benchmark functions and 12 real-world problems. The experimental results demonstrate that the proposed CMOEA outperforms existing CMOEAs in its convergence and diversity.

The effective resolution of constrained multi-objective optimization problems (CMOPs) requires a delicate balance between maximizing objectives and satisfying constraints. Previous studies have demonstrated that multi-swarm optimization models exhibit robust performance in CMOPs; however, their high computational resource demands can hinder convergence efficiency. This article proposes an environment selection model based on Bayes’ theorem, leveraging the advantages of dual populations. The model constructs prior knowledge using objective function values and constraint violation values, and then, it integrates this information to enhance selection processes. By dynamically adjusting the selection of the auxiliary population based on prior knowledge, the algorithm significantly improves its adaptability to various CMOPs. Additionally, a population size adjustment strategy is introduced to mitigate the computational burden of dual populations. By utilizing past prior knowledge to estimate the probability of function value changes, offspring allocation is dynamically adjusted, optimizing resource utilization. This adaptive adjustment prevents unnecessary computational waste during evolution, thereby enhancing both convergence and diversity. To validate the effectiveness of the proposed algorithm, comparative experiments were performed against seven constrained multi-objective optimization algorithms (CMOEAs) across three benchmark test sets and 12 real-world problems. The results show that the proposed algorithm outperforms the others in both convergence and diversity.

Achieving a dynamic equilibrium among feasibility, convergence, and diversity remains a fundamental challenge in Constrained Multi-objective Optimization Problems (CMOPs). To address the limitations of conventional methods in handling complex constraints and resource allocation, this paper proposes a Dual-Population Cooperative Evolutionary Algorithm based on Relaxed Feasibility Selection and Shrinking Contribution Resource Allocation (RFSCMOEA). First, a relaxed feasibility selection strategy is designed with a dynamically shrinking threshold, allowing near-feasible solutions to survive in early stages to enhance boundary exploration. Second, a dual-criterion environmental selection mechanism integrates non-dominated sorting with k-nearest neighbor density estimation to prevent premature convergence and ensure solution uniformity. Furthermore, a dynamic resource allocation model optimizes computational configuration by adjusting offspring generation ratios based on the real-time evolutionary contribution of each population. Extensive experiments on 47 benchmark functions and 12 real-world engineering problems demonstrate that RFSCMOEA significantly outperforms eight state-of-the-art algorithms in Feasibility Rate, Inverted Generational Distance, and Hypervolume.