Abstract:

Since current constrained optimization algorithms are easy to fall into the local optimum and their robustness are weak, a self-adaptive εconstrained optimization algorithm is proposed. By improving the individual comparison criterion, it can make full use of effective information carried by the infeasible solution, then enhance the exploration in the search space and improve the population diversity. Simultaneously, a self-adaptive adjustment strategy is presented to produce a suitable εfor balancing the relationship between the objective function and the constraint violation degree according to different problems, which can make more reasonable comparisons between individuals. Finally, comparative experiments on thirteen benchmark functions show that the proposed algorithm is not only able to converge the global optimal solution with higher accuracy but also has better robustness.