Infeasibility driven approach for bi-objective evolutionary optimization

Abstract

Infeasibility driven approach is proposed in this paper for constrained bi-objective optimization using evolutionary algorithm. The idea is motivated from one of the constraint handling techniques in which infeasible solutions are preserved in the population for focusing the optimal solution lying on the boundary of feasible region. In the proposed approach, extreme solutions of the current non-dominated front are allowed to recombine only with extreme infeasible solutions. This restricted mating is expected to generate offspring towards the “Paretooptimal” front and reduces number of generations required to evolve comparative results against existing multi-objective evolutionary algorithm (MOEA). Although the proposed approach is generic and can be coupled with any MOEA, but for bench-marking purpose it is coupled with NSGA-II (refer as IDMOEA) and is tested on four engineering optimization problems. On an average for 30 different runs, IDMOEA shows quicker convergence than NSGA-II with equivalent quality of solutions assessed by indicator analysis.