Biased multiobjective optimization for constrained single-objective evolutionary optimization

Constrained optimization problems widely exist in real world applications. A popular method to solve the problems is to transform the constraints to additional objectives and then the transformed problems can be solved by the multiobjective optimization techniques. The original multiobjective optimization techniques, such as Pareto dominance, lack the search bias to the constraints, which may lead to inferior performance of the algorithm. In this paper, first a concept of biased dominance or b-dominance to bias the search directly and explicitly is introduced, and then by combining b-dominance with a differential evolution, a Biased Multiobjective Optimization (BMO) algorithm is proposed for solving the constrained single-objective optimization problems. The BMO algorithm is comprehensively evaluated on 22 well-known benchmark test functions and the experimental results indicate that the BMO algorithm is capable of achieving competitive results compared with other state-of-the-art methods.