Benchmarking multiobjective evolutionary algorithms and constraint handling techniques on a real-world car structure design optimization benchmark problem
Hiroaki Fukumoto, Akira Oyama · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2018
While many of real-world industrial design problems involve several constraints, researches on multiobjective evolutionary algorithms (MOEAs) for problems with many constraints or the benchmark problems themselves are limited. The novel constrained multiobjective optimization benchmark problem based on a real-world car structure design optimization problem, termed Mazda CdMOBP, has more desirable characteristics as a constrained benchmark problem than the existing ones. The experimental results with 12 constrained MOEAs on this problem suggest the importance of balancing all of three factors of convergence, diversity, and feasibility and knowledge of proper settings of not only MOEA and CHT but also these parameters are imperative for application of MOEAs to industrial design problems.