Vehicle-cargo Matching For Maximizing Cargo Shipment Volume

The network freight platform enhances the ability of cargo owners and drivers to find goods through digital technology,effectively solves the problems of empty trucks and the backlog of goods,and is widely used in bulk logistics fields such as steel logistics.However,as the core function of the network freight platform,-cargo matching is still in the process of digital transformation and upgrading of the steel logistics industry faced with a series of problems: 1)Limited by the dynamic arrival of freight vehicles and the distribution of the number of trucks in different transportation directions for problems such as unevenness,only maximizing the transportation weight of vehicles arriving in real-time is the optimization goal,which would easily lead to the local optimization of vehicle-cargo matching results,that is,some vehicles have the largest transportation weight but the maximum volume of goods sent by the freight platform cannot be guaranteed.2)Given the weight difference between the load limit of a freight vehicle and the weight requirement for the goods shipped by the order,the existing vehicle-cargo matching strategy tends to produce more remaining orders(called tail orders)that are less than the weight of the vehicle),which need to add more vehicles to transport separately,greatly increasing freight costs.3)Since different types of goods require different types of freight vehicles to be transported,truck drivers tend to accept cargo transportation tasks familiar with the transportation flow,and drivers have transportation preferences for goods and routes,which brings severe challenges to cargo allocation and tail order consolidationIn practice,bulk freight companies mostly use combinatorial optimization methods or heuristics to determine the delivery scheme to maximize the load of real-time arriving vehicles,and this greedy strategy fails to consider the impact of the current vehiclecargo matching results on subsequent distribution decisions,resulting in the vehicle-cargo matching results falling into local optimum.In recent years,a small number of scholars have carried out research on vehicle-cargo matching methods for bulk logistics,and they mainly aim to maximize the weight of high-priority shipments and try to determine the length of the matching time window according to the number of vehicles arriving in different time periods to maximize the matching degree of trucks and goods.However,they did not pay attention to the minimum weight of the tail order cargoes that were stranded in the separate shipment,so the freight platform still could not achieve the goal of maximizing the total weight of the shipped cargo.Unlike online hiring platforms,ride-hailing dispatch is mainly based on the proximity between vehicles and passengers,while the matching of vehicles and goods in bulk logistics needs to be both Factors such as the flow of transportation and the type of cargo that can be carried by the driver are also considered.Therefore,the dispatch method in the travel platform cannot be directly used to solve the vehicle-cargo matching problem in the network freight platform.In view of the above problems,this paper takes steel logistics as an example and proposes a vehicle-cargo matching framework to maximize the total weight of goods shipped based on the analysis of multi-source data such as orders,vehicles,and waybills,including minimizing the distribution decision of remaining goods and maximizing the carrying capacity,so as to ensure the total weight of incoming cargo shipment is the largest in most of the time period.The main work of this paper includes the following:(1)Vehicle-cargo matching framework to maximize cargo shipments: By analyzing the historical data of real steel logistics enterprises,the sorting rules on the steel freight platform and the consolidation restrictions of vehicle loading are explored,and a vehicle-cargo matching framework is proposed to maximize the total weight of shipped goods.The vehicle-cargo matching process in steel freight is divided into two tasks,namely,the shipment of the whole vehicle of the waybill and the consolidation of the tail order.At the same time,the tail order problem caused by the existing vehicle-cargo matching method and the drawbacks of tail order consolidation are analyzed in depth,and the distribution method with the maximum shipping weight is designed globally combined with reinforcement learning technology,which consists of the order distribution decision based on Q-learning to minimize the remaining goods and the tail order consolidation method based on the deep Q network.(2)Order allocation method that minimizes remaining goods:A distribution method is designed to minimize the total number of remaining goods on the platform.Firstly,the goods in the order to be shipped are dynamically divided into multiple assembly tasks for truckload shipment according to the optimization goal of minimizing the remaining quantity of the tail order,and then the remaining tail order is packaged into the assembly task of consolidation shipment by using the branch demarcation method to minimize the remaining tail goods on the platform.In view of the local optimal matching problem,the gluttonous matching strategy is replaced by the binary graph matching method.In addition,considering the impact of the vehicle and cargo bipartite chart matching decision in the current period on the matching decision in the subsequent period,the Q-Learning algorithm is used to train the online binary graph matching strategy and use it to guide the online vehicle and cargo matching process to ensure that the global cargo shipment volume is maximized while the remaining number of goods on the platform is minimized.Experimental results on real data sets and synthetic data sets show that compared with other matching methods,our method can improve the cargo shipment optimization target by 8.7%.(3)Tail order consolidation method to maximize load capacity:A tail order consol-idation method is designed to maximize the total transportation weight of the consolidation vehicle.Firstly,aiming at the local optimal consolidation results caused by the greedy consolidation strategy,the tail order consolidation process with different transportation flow directions is regarded as a sequential decision problem,and the Markov process is modeled.At the same time,considering the influence of the final order consolidation decision in the current period on the decision of tail order consolidation in different flow directions in the future period,the definition of consolidation correlation degree is proposed,and on this basis,a tail order consolidation method based on the consolidation correlation degree is designed.By using the distribution of tail goods in different flow directions in the current time period to calculate the consolidation correlation tensor in the global space,in order to improve the consolidation efficiency,a pruning strategy based on the similarity of goods types is further proposed to prune the global consolidation correlation tensor calculation,and combine it with the tail order characteristic information in the current time period into the environmental state.Lastly,Use deep Q network training to obtain an online tail order consolidation strategy to ensure that the global tail order shipment volume is maximized.Through experiments on real data sets,the method proposed in this question can improve the optimization target of tail order shipment volume by 7.3%compared with the traditional greedy consolidation strategy.In this paper,a vehicle-cargo matching framework is designed to maximize the total weight of cargo shipment for the vehicle-cargo matching problem of steel freight platform,and the effectiveness and efficiency of the algorithm are verified through a large number of comparative experiments.Experimental results show that compared with the traditional greedy matching strategy,the framework proposed in this paper can achieve good results in different optimization goals: cargo shipment volume and tail order shipment volume.