Abstract: The multi-objective optimization problem aims to find a set of uniformly distributed Pareto optimal solutions. When there are constraints in these optimization problem, the Pareto optimal solution set often spreads across different feasible regions. For large-scale and tightly constrained problems, existing evolutionary algorithms have difficulty in balancing diversity and convergence of individuals, resulting in the search population possibly not being able to cross multiple infeasible regions and prematurely converging. To address this issue, an effective information-guided adaptive constrained multi-objective evolutionary algorithm is proposed. Firstly, this algorithm adopts a dual-population co-evolution mode. The main population (CA) mainly searches within the feasible region and converges to the Pareto front, while the auxiliary population (DA) retains infeasible solutions with smaller objective function values through environmental selection based on the constraint threshold, maintaining population diversity. Secondly, to improve the quality of individuals, a mating pool selection strategy based on the feasible solution ratio is designed. Finally, on 24 commonly used benchmark test problems, the proposed algorithm is compared with 5 similar approaches. The experimental results show that the algorithm proposed in the manuscript has certain advantages in handling nonlinear constrained multi-objective optimization problems.

Key words: multi-objective optimization problem, evolutionary algorithms, constraints, selection strategy, optimal solutions

摘要： 多目标优化问题旨在找到一组均匀分布的Pareto最优解，当问题存在约束限制时，Pareto最优解集往往分散在不同的可行区域。对于大规模紧约束问题，现有进化算法很难兼顾多样性和收敛性，导致搜索种群可能无法跨越多个不可行区域而早熟收敛。针对这个问题，提出了一种有效信息引导的自适应约束多目标进化算法。首先，该算法采取双种群协同进化模式，主种群(CA)主要在可行区域进行搜索并且收敛到Pareto前沿，而辅助种群(DA)通过基于约束阈值的环境选择，保留目标函数值小的不可行解，维持种群多样性。其次，为了改进种群个体质量，基于可行解比率指标设计了交配池选择策略。最后，在24个常用的基准测试问题上，将设计的算法与5种同类算法进行比较，实验结果表明，所设计的算法在处理非线性约束多目标优化问题时具有一定的优势。

关键词: 多目标优化问题, 进化算, 约束, 选择策略, 最优解