Research On Dual-Stage Constrained Multi-Objective Evolutionary Algorithm Based On Problem Type

Constrained multi-objective optimization problem(CMOP)is a common problem in various fields,and it has been widely used in practical problems such as engineering design,scheduling optimization,path planning and resource optimization.Therefore,solving constrained multi-objective optimization problems has very important scientific research value and practical significance.Solving CMOP should take into account the feasibility,diversity and convergence of the solutions in the population at the same time,so as to ensure that the population converges to a broad and evenly distributed Pareto front.Considering the above problems,this paper mainly carried out the following three parts of research work,and verified the convergence,diversity and stability of the algorithm through experiments.1.A new method for estimating the problem type in constrained multi-objective optimization is proposed,which estimates the problem type based on the proportion of feasible solutions in the main population at the end of the first stage.According to the experimental results,the proposed method has a high estimation accuracy.2.A dual-stage constrained multi-objective evolutionary algorithm based on problem type(DT-CMOEA)for solving constrained multi-objective optimization problems is proposed.This algorithm divides the evolution process into two stages.The first stage is to find the corresponding UPF.At this time,only the evolution of the objective function is considered,so that the population can quickly cross the infeasible domain and maximize the search range of the population in the target space.In the process of evolution,the nondominated feasible solution of each generation is preserved to realize the diversity of the population.The second stage mainly includes three parts: problem type estimation,population update and evolutionary algorithm matching.This stage mainly considers the feasibility and convergence of the solutions in the population,so that the solutions are finally widely and evenly distributed along the constrained Pareto front.3.Aiming at the uneven distribution of offspring solutions that may occur in the process of environment selection,this paper proposes an improved constrained multi-objective evolutionary algorithm based on problem type(IDT＿CMOEA),which improves the generation and selection process of offspring solutions.The algorithm uses an adaptive parameter adjustment strategy to improve the performance of the DE algorithm and generate descendant solutions with better properties.Then,in the selection process of descendant solutions,first determine the non-dominated area,then use the weight vector to divide the non-dominated area,and select the optimal solution for each sub-area after division,so that the solutions selected to participate in the subsequent evolution process are more widely distributed in the non-dominated area.The value is measured by a distance function based on the cost value,which ensures that the solution selected to participate in the subsequent evolution process has better properties.