Multi-objective Chaotic Optimization Algorithm by Combining Gray and Real Codes

How to avoid premature convergence and maintain diversity of Pareto-optimal solutions for multi-objective optimization problems is addressed. An improved non-dominated sorting genetic algorithm is put forward, in which Gray and real codes are used for crossover and mutation operations, respectively, and chaotic variables are applied to produce initial population and control the crossover/mutation operations. The Gray-code based crossover operator has greater search ability and can produce more new individuals which preserve the best genes. The mutation operator with real-code mechanism has larger search range and higher precision, which can enlarge the diversity of population. The proposed method is demonstrated using three classical test functions.