Research on Dynamic Logistics Transfer Contact Method Based on Reinforcement Learning
The unreasonable layout of the cigarette logistics transit network will cause many problems such as long delivery miles to terminals,long travel time,low loading rates and short customer service time.This paper applies reinforcement learning,clustering and center of gravity siting algorithm to sut up a dynamic cigarette logistics transit docking point optimization model to op-timize the cigarette logistics transit network.According to the number of orders and retail customers,delivery mileage,travel time and some other factors,it proposes a method to scientifically calculate and decides the dynamic transit docking points in the transit areas based on historical data training model,"Drop box"transportation model and the station resources in the tobacco commercial system.It uses the proposed model and algorithm to do the simulations over the data from Zhushan transit area of Hubei Tobacco Company Shiyan Branch.The experimental results show that the method with two transit points is optimal during the low season of cigarette sales,while the method with three transit points is optimal during the high season of cigarette sales.It also verifies the per-formance through actual vehicles in Danjiang transit area of Shiyan that during the low season of cigarette sales.The method with 2 transit points reduces the average terminal delivery time by 37%,the average terminal delivery mileage by 25.8%,the average trunk delivery time by 22.8%and the average trunk delivery mileage by 49.2%,compared to the original method with 1 transit point.These two transit areas alone are forecast to save Shiyan Tobacco approximately RMB 110,000 to 146,000 yuan in annual lo-gistics operating costs.This paper investigates the optimization model and method of dynamic cigarette logistics transit docking points to achieve the optimization objectives of reducing operating costs in transit area and terminal travel time,which can effective-ly improve the efficiency of cigarette logistics.
logistics distributionreinforcement learningdynamic cigarette transit"Drop box"transport model
国家科技期刊平台
NSTL
万方数据
