Research On Multi-objective Genetic Algorithm And Non-dominated Sort

Evolutionary algorithm is a kind of random algorithms based on natural selection and genetic, which characteristics are search strategies and message exchanges between different individuals. Especially, it can deal with the complex and difficult problems those traditional search methods couldn't solve. It Widely used in combinatorial optimization, machine learning, adaptive control, intelligent decision systems and artificial life and so on. With the development of the multi-objective evolutionary algorithms, the efficiency and applicability becomes popular in which the construction of the non-dominated set having more Influence.This paper mainly have the following areas:1) Introduce the origin of the genetic algorithm,the current evolutionary multi-objective algorithms and the basic concepts,terminology,frame and the genetic characteristics and trends.2) Analysis the traditional multi-objective optimization problems and NSGA-Ⅱ, showing the shortcomings in solving practical problems.3) Based on traditional evolutionary algorithms and NSGAII, we introduce an Arena Principle, which uses the individual comparison in construction of the non-dominated set. We do not need to sort the population in the procedure of computation. Two function tests show that the new algorithms based on Arena Principle and NSGAII sharply reduce the complex of the computation, getting the Distribution more Uniform.4) Based on Arena Principle and NSGAⅡ, in order to get the computation more better we introduce theε-Pareto to improve the genetic operators. Using that it can efficiently maintaining the population diversity,getting out of the local optimal solution and strengthening the ability of getting global optimal solutions.5) We use the classic experimental function tests, the results show that the ability of maintaining the population diversity strengthen and the dependence of the real problems weak.