Clustering Based Multiobjective Evolutionary Algorithm And Appliance In Route Planning

In daily life, people can always meet with many complicated multiobjective optimization problems(MOPs) with properties of multiple constrains, multiple variables and high non-linearity. Since the traditional deterministic optimization techniques cannot deal with these problems very well, the evolutionary algorithm(EA), which is inspired by natural evolution, has received great concentration and favorite. The solving process of multiobjecctive evolutionary algorithm(MOEA) is independent of the features of the problem to solve, and a MOEA is able to obtain plenty of solutions to be chose from. Therefore, the MOEAs have experienced rapid development.First of all, this paper delivered a brief introduction of the history and development of multiobjective evolutionary algorithm, and pointed out some major trend of future research. One of them is trying to improve the performance of MOEA by combining with the machine learning technology, and one another is to integrate more than only one reproduction operators in a MOEA and try to obtain a evolution algorithm with better solving capability.This paper proposed a clustering based multiobjective evolutionary algorithm(CMO). This algorithm builds a local Gaussian model for every solution, based on the consequence of K-means clustering. And CMO generate new solutions by applying sampling technique to the established Gaussian model. By this way, the algorithm achieved the purpose of searching in the decision space. Experiments and comparisons were conducted to get a view of the problem solving ability of proposed algorithm.Experimental analysis indicates that proposed algorithm has better ability compared to some selected algorithms. Sensitivity analysis was executed to CMO.Based on the CMO, the paper proposed a clustering based multi-operator multiobjective evoltionary algorithm(CMMO), which has been integrated with three different reproduction operators, and implanted with adaptive mechanism. The adaptive mechanism allows the CMMO to automatically alter the selection possibility of each reproduction operator according to the demands of algorithm in different stage. CMMO received a promising performance in comparison experiments. Solving ability of CMMO to other complicated multiobjective problems also has been certificated.Last part of this paper gives a general picture of the model of route planning for aircraft, and tried to solving the route planning problem using those proposed evolutionary algorithms. Analysis of performance and reasons has been conducted. And in the end, some better routes were given. The solving ability of proposed algorithms to complicated problems is proved.