Impact of Reformulation for the Real-World Constrained Multi-Objective Problems on the Performance of Evolutionary Algorithms

The real-world constrained multi-objective optimization problems (CMOPs) pose significant challenges for evolutionary multi-objective optimization (EMO) algorithms due to complex constraints. This paper explores the impact of reformulation strategies on EMO algorithms, focusing on equality constraint substitution and cumulative variable introduction. Experiments on 21 real-world CMOPs and their reformulated versions using NSGA-II, CCMO, and PPS-MOEA/D show that reformulation enhances feasibility rates and solution quality, particularly for low-feasibility and high-dimensional problems. Reformulating problems simplifies the search space, allowing algorithms like NSGA-II and CCMO to achieve significantly higher feasibility rates. These results underscore the effectiveness of problem reformulation in alleviating the difficulty of solving real-world CMOPs and provide valuable insights for improving EMO algorithm performance on complex problems.