A New Constrained Multiobjective Optimization Algorithm Based on Artificial Immune Systems

This paper proposes a new constrained multiobjective optimization algorithm based on artificial immune systems (AIS). To deal with constrained multiobjective optimization problems, the constrained AlS-based multiobjective optimization algorithm is developed by integrating a proposed constraint-handling technique with the unconstrained AIS-based multiobjective optimization algorithm named MOAIS (Xiao and Zu, 2006). We propose the constraint-handling technique by extending a single-objective constraint-handling technique called stochastic ranking (Runarsson and Yao, 2000) to multiobjective optimization process. Two scenarios of the multiobjective version of stochastic ranking are suggested. Thereafter, we develop the constrained MOAIS named MOAIS+SR by integrating the two scenarios with MOAIS. A comparative study is performed quantitatively to assess the performance of MOAIS+SR on a constrained test function suite called CTP test problems. In the comparative study, MOAIS+SR is compared against two other constrained multiobjective algorithms. The simulation results show that the proposed multiobjective stochastic ranking outperforms the constrained-dominance principle (Deb et al., 2000) in handling constraints. Furthermore, we show that the proposed MOAIS+SR achieves the best overall performance among the three algorithms under consideration on the CTP test problems. This study demonstrates that the proposed MOAIS+SR is highly competitive with other state-of-the-art algorithms in constrained multiobjective optimization.