Research On MOEA/D Based On Membrane Algorithm In Vehicle Routing Problem With Time Windows

Vehicle Routing Problem(VRP)is a common Problem in real life,which exists in many fields,such as logistics distribution management and supply chain.This thesis mainly studies VRP with Time Windows(VRPTW).With the minimum number of transport vehicles and the minimum total distance traveled by the vehicles as the two objectives,and vehicle path planning as an independent variable,the Pareto set composed of non-dominated solutions is solved to provide decision support for decision makers.VRPTW is a classic NP-Hard problem,and its problem scale is often beyond the scope of application of algorithms such as precise algorithms.Intelligent optimization algorithm has become the commonly used algorithm to solve such problems because of its excellent global search ability and fast convergence speed.In this thesis,the multi-objective evolutionary algorithm based on decomposition and membrane computing for solving multi-objective VRPTW is studied.The key work includes:(1)For multi-objective VRPTW,this thesis firstly chooses a multi-objective evolutionary algorithm based on decomposition(MOEA/D)with fast convergence speed,and adds appropriate local search operators on the basis of its strong global search ability to make up for the lack of local search capabilities.Firstly,the storage structures of the solution are set as chain storage and sequential storage,and the Tchebycheff approach is adopted in the neighborhood with heuristic vehicle number optimization operator and heuristic distance optimization operator respectively.Then,the evolution operation between neighborhoods is added to avoid falling into the local optimal solution.(2)Through the experimental analysis of the improved MOEA/D,it can be seen that it still has the shortcoming that the Pareto front boundary is not optimal.The solution in the single-dimensional still has room for optimization,so it is necessary to explore the Pareto front boundary.To solve this shortcoming,this thesis proposes a membrane algorithm based on MOEA/D and tissue-like P system(T-MOEA/D)for solving multi-objective VRPTW.By using the single-objective solution and the parallelism of P system,we can optimize the boundary and improve the computing power without increasing the time cost.In this algorithm,a tissue-like P system with four membranes is constructed.The two single-objective solutions of the bi-objective problem in this thesis,the operation of the improved MOEA/D and the self-searching of the external population are respectively set as the reaction rules in each membrane.Cell communication rules are used to transfer and share information to the objects in the system to ensure the stable operation of parallel evolution.Through simulation experiments and data comparison,it can be verified that the improved MOEA/D is feasible and effective,and its precision has certain competitiveness among the similar evolutionary algorithms.On this basis,T-MOEA/D further optimizes the distribution of Pareto set.