A Multi-Populated Differential Evolution Algorithm for Solving Constrained Optimization Problem

• Published in: 2006 IEEE International Conference on Evolutionary Computation pp. 33 - 40
• Main Authors: Tasgetiren, M.F., Suganthan, P.N.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 2006
• Subjects: Benchmark testing; Biological cells; Constraint optimization; Evolutionary computation; Genetic mutations; Model driven engineering; Optimization methods; Statistics; Stochastic processes; Upper bound
• ISBN: 9780780394872; 0780394879
• ISSN: 1089-778X
• DOI: 10.1109/CEC.2006.1688287

This paper presents a multi-populated differential evolution algorithm to solve real-parameter constrained optimization problems. The notion of the "near feasibility threshold" is employed in the proposed algorithm to penalize the infeasible solutions. The algorithm was tested using benchmark instances in Congress on Evolutionary Computation 2006. For these benchmark problems, the problem definition file, codes and evaluation criteria are available in http://www.ntu.edu.sg/home/EPNSugan. The performance of the multi-populated differential evolution algorithm is evaluated with the best known or optimal solutions provided in the literature. The experimental results with detailed statistics required for this session show that the proposed multi-populated differential algorithm was able to solve 22 out of 24 benchmark instances to either optimality or best known solutions in the literature. In addition, 6 out of 24 best known solutions are ultimately improved by the proposed multi-populated differential evolution algorithm.