Research On Truck Scheduling Problem Based On Multiple Cross-docking Doors

Third-party logistics,such as distribution centers,are committed to developing new operational methods to reduce operating costs and improve customer service level.Cross docking is a new strategy based on inventory management,without warehousing function.While arriving at a distribution center,the goods need to be sorted and stored temporarily,then re-integrated and delivered to the customers due to demands.Ephemeral storage of goods can reduce the inventory holding cost,the probability of damage and save space.Cross docking can improve the efficiency of distribution,and reduce logistics management costs.The facilities are limited in cross-docking centers,so when several trucks are ready for operation at the same time,the phenomenon of trucks waiting occurs.Therefore,truck scheduling is a key to reduce waiting time,which is even a key point of the cross docking.Truck scheduling is a NP problem,which allocates tasks for doors and lines up for trucks to reduce the total operation time.In this paper,the model of truck scheduling problem in multiple-door cross-docking system was established,due to the number of cross-docking doors.In order to minimize the operation time,searched for the best task assignment for doors and vehicle queues.Three heuristic algorithms,namely harmony search algorithm(HS),improved harmony search algorithm(IHS)and genetic algorithm(GA),were proposed to solve the MILP model.Aiming at the limitation that values of parameters have great influence on the searching ability of HS algorithm in different solving stages,IHS algorithm adopted a dynamic adjustment method to set parameters.Heuristic algorithm with different parameter values performed differently.The parameters values of each algorithm were optimized by Taguchi experiment to determine the optimal combination of parameters.Finally,according to the number of goods,categories and trucks,nine groups of experiments were designed from three dimensions of quantity(large,medium and small).The performances of the models and algorithms were analyzed from the perspectives of optimal solution,worst solution,average solution and CPU operation time.