Research On Uncertain Logistics Transportation Scheduling Problem Based On Deep Reinforcement Learning

Transportation occupies a very important position in the entire logistics industry.The cost accounts for 35%-50% of the total logistics cost and 4%-10% of the commodity price.Controlling the transportation cost plays an important role in saving the total logistics cost.With the continuous development of information technology,new technologies should be widely used in the logistics industry.The use of modern information technology can improve the logistics dispatch management capabilities of logistics enterprises,improve the service level of the logistics industry,reduce transportation costs,and thereby improve the competitiveness of logistics enterprises.For the uncertain logistics transportation scheduling problem,the current solution mainly has the following three major problems: first,the use of emerging technologies such as artificial intelligence and big data is still lagging behind;second,it is often used when solving certain logistics transportation scheduling problems The heuristic intelligent optimization algorithm is not suitable for uncertain logistics transportation scheduling problems.Third,uncertain logistics transportation scheduling models established for using traditional heuristic algorithms often have many constraints and are very complicated.This paper is devoted to research and application of a new type of intelligent algorithm(Deep Reinforcement Learning Algorithm)in uncertain logistics transportation scheduling system.From the perspective of the economic benefits of logistics transportation enterprises,multiple vehicle routing problem models are established according to the actual situation of uncertain logistics transportation scheduling problems,and the common deep reinforcement learning algorithm and its improved algorithm are used to carry out the uncertain logistics transportation scheduling model one by one.Solve.The main work of this article is as follows:(1)Study the uncertain logistics dispatching problem of a single distribution center and establish a model.It is proposed to use Deep Q Network based on Pointer Network to solve the model,and a feasible logistics transportation scheduling strategy can beobtained through simulation experiments.Simulation results show that the algorithm is superior to other common intelligent optimization algorithms in terms of solving accuracy and stability,which proves that the algorithm is effective and feasible in solving uncertain logistics transportation scheduling problems.(2)Study the uncertain logistics transportation scheduling problem of multiple vehicle models and establish a model.The Pointer Network is also used as a method of deep learning,and a REINFORCE algorithm with a baseline is proposed to solve the model.Through simulation experiments,a feasible logistics transportation scheduling strategy can be obtained.The simulation results show that the REINFORCE algorithm with baseline can solve the multi-vehicle uncertain logistics transportation scheduling problem with a higher accuracy,which proves to be a method with practical application value.(3)Study the uncertain logistics transportation scheduling problem with time window and establish a model.An Actor-Critic algorithm based on Pointer Network is proposed to solve the model.Through simulation experiments,feasible logistics transportation scheduling strategies can be obtained.Finally,through experimental simulations,it is proved that the pointer network-based actor-reviewer calculation has a good effect on solving the uncertain logistics transportation scheduling problem with time windows.(4)Study uncertain logistics transportation scheduling problems in multi-distribution centers and establish models.An Actor-Critic algorithm based on improved pointer network is proposed to solve the model.Through the modification of the pointer network model,it is more suitable for the model with more complex input established in this chapter.Through simulation experiments,feasible logistics transportation scheduling strategies can be obtained.The final experimental simulation results show that the Actor-Critic algorithm based on the improved pointer network can greatly improve the efficiency of the algorithm under complex input conditions,which proves to be an effective improvement.The innovation of this paper is to establish a mathematical model of uncertain logistics transportation scheduling problem suitable for deep reinforcement learningalgorithm,and use Pointer Network as Deep Q Network,REINFORCE algorithm with baseline and Actor-Critic algorithm and other deep reinforcement learning algorithms' Deep learning network to solve uncertain logistics transportation scheduling problems.Experiments show that all the methods proposed in this paper get good results.