Research Of Multi-objective Evolutionary Algorithm Based On Decomposition And Its Application In Path Planning

Multi-objective optimization problem exists widely in engineering practice,these multi-objective optimization problems have conflict and coupling among objectives.However,the existing evolutionary algori thms based on decomposition still need to be improved in balancing convergence and diversity.Moreover,the traditional evolutionary algorithm can not solve the optimization problem with complex and irregular Pareto fronts.In this paper,the optimization problem of equilibrium convergence and diversity as well as complex and irregular Pareto front s is studied,and it is applied to practical problems su ch as robot path planning.The multi-objective optimization algorithm based on dominance could reduce the selection pressure when solving high-dimensional problems because of the non-domination of each object.However,the multi-objective optimization algorithm based on decomposition decomposes the multi-objective optimization problem into a group of single objective optimization problems according to a group of weight vectors,which effectively solved this kind of problems and could effectively improve the distribution of solutions.But the diversity of solutions is still insufficient.Therefore,in order to balance the convergence and diversity,a decomposition multi-objective optimization algorithm based on the minimum distance and aggregation strategy is proposed.Firstly,for the sake of improving the diversity of population,the algorithm used the angle decomposition technology to decompose the target space into specified number of subspaces.What's more,it used the decomposition technology based on minimum distance and aggregation strategy in two stages in each subspace.In this strategy,the minimum distance was used to select the soluteons to improve the convergence,and then the aggregation strategy was used to improve the distribution of solutions.In the process of generating new solutions,the aggregation based cross neighborhood method was added to make the new solutions generated more retain the characteristics of the optimal solutions of the previous generation.Finally,compared the proposed algorithm with the classic algorithm on the ZDT and DTLZ series of test functions.Experimental results show that the proposed algorithm can significantly improve the convergence and diversity of the solution set.Concerning the issue that the low quality of offspring solutions generated by traditional evolutionary operators,and the fact that the evolutionary algorithm based on decomposition can't solve the multi-objective optimization problem with complicated Pareto fronts,a multi-objective evolutionary algorithm based on decomposition use of double reference point and prediction by historical information was proposed.Firstly,the evolutionary operator based on historical information prediction was used to generate better offspring solutions to improve the convergence of the algorithm;secondly,the decomposition strategy based on ideal point and nadir point was used to select solutions to solve the optimization problem with complicated Pareto fronts,and the decomposing method with augmentation term was used to improve the population diversity when selecting solutions according to the nadir point.Finally,for the sake of verify the feasibility of the algorithm,the F-series test problem with complicated Pareto fronts were used to test.It has been verified that the overall performance of the proposed algorithm is superior to that of the comparison algorithm.The evolutionary operator based on historical information prediction can produce better quality offspring solutions,and the decomposition strategy based on double reference points can well solve the multi-objective optimization problem with complicated Pareto fronts.In practical engineering,robots are generally in a complex and dangerous environment.How to make robots avoid obstacles and dangerous sources and find an optimal path is the focus of robot motion research.Therefore,for the sake of solve the problem of robot path planning in the environment of obstacles and hazards,the above two evolutionary algorithms are improved,and two improved path planning algorithms based on multi-objective evolution are proposed.The two algorithms introduce the local shortest path solution in the initialization process and the global optimal guidance set in the evolution process,and combine the global optimal guidance set with the evolution operator to i mprove the search ability of the algorithm.By setting two different simulation environments with obstacles and hazards,compare and analyze the proposed path planning algorithm with the path planning algorithm based on traditional multi-objective evolution,which proves the effectiveness and advantages of the proposed algorithm in dealing with robot path planning problems in an environment with obstacles and hazards.