Abstract:

Concerning the high time-complexity of multi-objective evolutionary algorithm based on Pareto dominated solution sorting, considering the potential features of non-dominated solution sorting, a fast sorting method which only handles individuals with the highest rank in current population was introduced. The individuals could be chosen into the next generation during the sorting. The algorithm was terminated when the population of next generation was selected enough, which reduced the number of individuals for sorting process and the time complexity. In addition, a method of uniform crowding distance calculation was given. Finally, a Fast Multi-Objective Differential Evolution (FMODE) algorithm was proposed which incorporated the introduced sorting method and uniform crowding distance into Differential Evolution (DE). Simulation experiments were conducted on the standard multi-objective optimization problems including ZDTl~ZDT4 and ZDT6. The time consumption of FMODE was far less than NSGAⅡ in large size of population, and its overall performance was much better than the classic NSGAⅡ, SPEAⅡ and DEMO. Moreover, the uniform crowding distance method provided better performance than classic crowding distance in framework of FMODE. The parameters of FMODE were also obtained by experiments. The simulation results show that the proposed algorithm has lower time complexity of calculating level, and better performance in convergence and diversity.