Dynamic Escaping Population Driven Evolutionary Algorithm for Multi-objective Optimization Problems with Deceptive Constraints
提出一种动态逃逸种群驱动的进化算法DEPEA，通过三个协同种群和边界多样性增强策略，有效解决约束违反值不可靠的欺骗性约束多目标优化问题，在基准测试和实际问题上优于13种现有算法。
Deceptive constrained multi-objective optimization problems (DCMOPs) pose severe challenges to constrained multi-objective evolutionary algorithms (CMOEAs), as constraint violation (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">CV</i>) values become unreliable. In this paper, a dynamic escaping population-driven CMOEA, named DEPEA, is proposed by maintaining three cooperative populations: an escaping population (EP), an auxiliary population (AP), and a main population (MP). The EP employs slow and rapid escape modes to regulate its escaping behavior at different evolutionary stages, the AP aims to converge toward the unconstrained Pareto front (UPF), and the MP focuses on approximating the constrained Pareto front (CPF). To avoid mutual interference between the AP and EP while controlling the escaping behavior of the EP, a one-way fine-grained cooperation mechanism is designed. In addition, a boundary diversity enhancement strategy is proposed to strengthen exploration along boundary regions. Experimental studies conducted on four benchmark suites and six real-world constrained optimization problems demonstrate that DEPEA achieves superior performance compared with thirteen state-of-the-art CMOEAs.