An Annealing Stochastic Ranking Mechanism for Constrained Evolutionary Optimization
Weiqin Ying, Dongxin Peng, Yuehong Xie, Yu Wu · 2016
Stochastic ranking is a promising constraint handling method for constrained evolutionary optimization problems. However, the method has some limitations when comparing two given individuals during the evolution. In order to find the optimal feasible solution, the promising infeasible individual is supposed to have an opportunity to survive if at least one of two compared individuals is infeasible. In the paper, an annealing stochastic ranking mechanism is proposed by borrowing the idea of the Metropolis acceptance criterion in simulated annealing. The mechanism is able to take full advantage of promising infeasible solutions to guide the search towards the global optimal feasible solution during the evolution. Experimental studies on 13 benchmark problems demonstrate that the annealing stochastic ranking mechanism improves the search performance of the classic stochastic ranking significantly.