AGENT-BASED EVOLUTIONARY ALGORITHMS APPLIED TO CONSTRAINED MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

摘要

Traditionally, constrained multi-objective optimization problems are difficult and are rarely dealt with by agent-based evolutionary algorithms. In response to the difficulties, a compatible agent-based evolutionary algorithm is introduced in which the normalized degree of the violation of the constraints is considered as an additive objective able to influence the energy of the agents. In addition, an inclusion of two metrics is adapted to solve the intractable problem. Initially, two external archives—optimal solution set and feasible optimal solution set—are available to maintain the diversity of the population. Then, an ad hoc climbing operator is suggested to promote both candidate solutions and agents with small violation degrees to achieve feasible optimal solutions efficiently. Case studies consisting of four testing functions show that the proposed algorithm not only keeps the diversity of the population but also converges to the optimal fronts quickly.