Research On Optimization Model And Improved TLBO Algorithm For Logistics Distribution Path Planning

Reasonable planning of logistics distribution paths is of great significance for reducing logistics transportation costs and improving logistics distribution efficiency.Logistics distribution path planning is essentially a vehicle routing problem(VRP).This thesis focuses on the models and solving methods of three variants of VRP,which are capacitated VRP(CVRP),green VRP(GVRP),and GVRP with fuzzy demand(GVRPFD).An improved discrete TLBO with local search(IDTLBO-LS)algorithm is proposed and applied to CVRP.Considering the shortcoming of TLBO algorithm which is easy to fall into local optima and the CVRP,we improve and extend TLBO algorithm from four aspects: population initialization,discretization of TLBO algorithm,combination with local search(LS)algorithm and adding population reconstruction mechanism.Firstly,a hybrid initialization method is designed to improve the quality of the initial population by integrating random initialization and nearest neighbor initialization.Then,an improved discrete teaching-learning strategy is designed,and two crossover operators are adopted to simulate teaching phase and learning phase respectively.Then,a LS algorithm is designed,and it is combined with the improved discrete TLBO(IDTLBO)algorithm to further improve the quality of solutions.Finally,a population reconstruction mechanism is designed to make the population jump out of local optima and avoid premature convergence.Simulation experiments show that compared with traditional discrete TLBO(DTLBO)algorithm,DTLBO with LS(DTLBO-LS)algorithm,LS algorithm,IDTLBO algorithm and IDTLBO-LS algorithm with random initialization(IDTLBO-LS-R),IDTLBO-LS algorithm converges faster and has the highest convergence accuracy.Compared with recently proposed representative algorithms such as hybrid genetic and simulated annealing(HGSA),improved flower pollination algorithm(IFPA)and improved hybrid firefly algorithm(CVRP-FA),IDTLBO-LS algorithm has the highest convergence accuracy.On the basis of CVRP,an improved discrete TLBO algorithm for GVRP is proposed.Firstly,the mathematical model of GVRP is established,which incorporates the objective of fuel consumption reduction in the objective function of CVRP.Then the initialization method and local search algorithm in CVRP’s IDTLBO-LS algorithm are redesigned,and a new IDTLBO-LS-G algorithm is proposed to make it applicable for GVRP.Simulation experiments show that IDTLBO-LS-G algorithm has the highest convergence accuracy compared with the traditional DTLBO algorithm,the poposed IDTLBO algorithm,the DTLBO-LS algorithm and LS algorithm.Compared with the IDTLBO-LS algorithm of CVRP,the IDTLBO-LS-G algorithm is more suitable for solving GVRP and has higher convergence accuracy.On the basis of GVRP,the fuzziness of customer demand and decision-maker preference is further considered,and the mathematical model of GVRPFD is established and its solving method is designed.The description,aggregation and propagation methods of fuzzy information in GVRPFD are deeply concerned,solving the difficulties inherent in the complexity,dynamics,fuzziness,uncertainties of the GVRPFD.Firstly,considering the uncertainty of decision making process and decision environment,the processing method of fuzzy demand and fuzzy weights by using gaussian interval type-2 fuzzy sets is proposed,and the fuzzy demand constraint is processed.Then,the perceptual compuing theory and linguistic weighted average operator are used to aggregate the uncertain information.Then,according to the mathematical model of GVRPFD,GVRP’s IDTLBO-LS-G algorithm is adopted to solve the GVRPFD.Simulation results show that compared with the type-1 fuzzy set-based GVRPFD mathematical model and the simple weighted average-based GVRPFD mathematical model,the type-2 fuzzy set-based GVRPFD mathematical model proposed in this thesis can describe and deal with the uncertainty of customer demand and decision maker preference more reasonably,and can give more accurate and reliable optimization results.