Vehicle-Cargo Matching Method and System for Multi-Node Mode

Abstract

The present invention discloses a vehicle-cargo matching method and system for a multi-node mode, which relates to the field of logistics operation management. The present invention first uses a two-stage strategy of clustering based on the geographical coordinate positions of cargo destinations and then path planning, making full use of the geographical location information of the cargo and saving the solution time of the problem. The algorithm first uses the silhouette coefficient method to determine the clustering number k value, and applies the k-means algorithm to generate a cargo clustering scheme. Subsequently, a heuristic algorithm based on the matching fitness is used to match the cargo clusters with the vehicles. Then, for the path planning of each vehicle, the transfer probability design of the ant colony algorithm is improved, the distance and time factors are comprehensively considered, and an adaptive evaporation factor and a local search strategy are adopted, so as to realize the optimal planning of the vehicle path, improve the transportation efficiency and cost savings.

Description

Technical Field

[0001] The present invention relates to the field of logistics operation management, and more specifically to a vehicle-cargo matching method and system for a multi-node mode. Background Art

[0002] According to relevant research progress at home and abroad, although a large number of research results have been achieved in the fields involved in vehicle-cargo matching, there are still certain limitations in the application of existing methods in vehicle-cargo matching. By comparing the research results of predecessors, the following main deficiencies can be summarized:

[0003] The vehicle-cargo matching problem involves various scenarios in practice, such as assembly and transshipment, etc. However, most of the existing literature only considers simple one-to-one scenarios. Due to the limited application scenarios, there are relatively few studies on cargo assembly or transshipment transportation involving one-to-many, and more in-depth research is needed. Summary of the Invention

[0004] In view of this, the present invention provides a vehicle-cargo matching method and system for a multi-node mode, which adopts a two-stage strategy of first clustering according to the geographical coordinate positions of cargo destinations and then path planning, making full use of the cargo geographical location information and saving the problem-solving time. The algorithm first uses the silhouette coefficient method to determine the clustering number k value, and applies the k-means algorithm to generate a cargo clustering scheme. Subsequently, a heuristic algorithm based on matching fitness is used to match the cargo clusters with vehicles. Then, for the path planning of each vehicle, the transfer probability design of the ant colony algorithm is improved, considering both distance and time factors, and an adaptive evaporation factor and a local search strategy are adopted to optimize the vehicle path planning, improving the transportation efficiency and cost savings.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A vehicle-cargo matching method for a multi-node mode includes the following steps:

[0007] Determine the number of clusters according to the freight information and vehicle information, and use the k-means algorithm combined with the silhouette coefficient method to cluster the freight information to obtain cargo clusters;

[0008] Match each cargo cluster with the vehicle information based on heuristic rules;

[0009] Use a multi-strategy improved ant colony path planning algorithm to plan the vehicle path and output the optimal solution of vehicle-cargo matching.

[0010] Optionally, the steps of determining the clustering cluster value k by the silhouette coefficient method are as follows:

[0011] Run K-means clustering: run each best k-means candidate value selected by the elbow method;

[0012] Calculate the silhouette coefficient: For each data point, calculate its silhouette coefficient:

[0013]

[0014] where a represents the average distance from the data point to other points in the same cluster, and b represents the average distance from the data point to the nearest cluster;

[0015] Calculate the average silhouette coefficient: Calculate the average silhouette coefficient corresponding to each optimal k-means candidate value, that is, take the average of the silhouette coefficients of all data points;

[0016] Draw the average silhouette coefficient curve: Plot the average silhouette coefficient of each K value as a curve graph. Observe the curve, the silhouette coefficient ranges between [-1, 1], and the larger this value, the more reasonable the classification, so as to determine the K value.

[0017] Optionally, match each cargo cluster with vehicle information based on heuristic rules. The specific steps are as follows:

[0018] Determine the cargo volume of each cluster: Calculate the total weight and total volume of all cargos in each cluster, and compare with the vehicle models with the maximum load capacity. If it exceeds, eliminate the customer farthest from the clustering center until the capacity limit is met;

[0019] Cargo reallocation: Determine whether there is excess vehicle capacity in other customer groups. If so, incorporate the eliminated customers into the nearest customer group. If not, form a separate cluster;

[0020] Calculate the adaptability of each vehicle model. The calculation formula is as follows:

[0021]

[0022] where w i represents the total weight of the i-th cluster, W k represents the load of the k-th vehicle, and δ ik represents the fitness value of the k-th vehicle relative to the i-th cluster;

[0023] Use the greedy algorithm to allocate vehicles one by one in the order of adaptability from high to low.

[0024] Optionally, the specific steps of the multi-strategy improved ant colony path planning algorithm are as follows:

[0025] Initialize the parameters and set the maximum number of iterations iter max ;

[0026] Based on the transition probability Calculate the jump probability of accessible nodes, use the roulette wheel strategy to select the jump node of the ant, and update the taboo list tabulist k, add the selected cities to it, and repeat the operation until a complete path is constructed, that is, all cities are visited and returned to the starting city;

[0027] Calculate the cost of each path, record the current optimal cost and the corresponding path;

[0028] Calculate the adaptive evaporation coefficient and update the pheromone concentration τ ij (t + 1);

[0029] If the maximum number of iterations iter is reached max , the algorithm stops running and outputs the optimal result; if the maximum number of iterations iter is not reached max , clear the taboo list of all ants and continue the iteration;

[0030] Output the best path found and its corresponding objective function value.

[0031] Optionally, the transition probability is calculated as follows:

[0032]

[0033] where n ij (t) represents the heuristic visibility, α represents the time penalty coefficient, β represents the expected heuristic factor, and τ ij (t) represents the pheromone.

[0034] Optionally, the expression formula of the adaptive evaporation coefficient is as follows:

[0035]

[0036] where L mean represents the average length of all paths in the current generation; L min represents the shortest length among all paths in the current generation.

[0037] A vehicle-cargo matching system for a multi-node mode, including:

[0038] Cargo clustering module: used to determine the number of clusters according to freight information and vehicle information, and use the k-means algorithm combined with the silhouette coefficient method to cluster the freight information to obtain cargo clusters;

[0039] Vehicle-cargo matching module: used to match each cargo cluster with vehicle information based on heuristic rules;

[0040] Optimal path planning module: used to plan the vehicle path using a multi-strategy improved ant colony path planning algorithm and output the optimal solution of vehicle-cargo matching.

[0041] As can be seen from the above technical solutions, compared with the prior art, the present invention provides a vehicle-cargo matching method and system for a multi-node mode. Combining with actual transportation constraints, an optimization model for minimizing transportation costs is established, and an ant colony vehicle scheduling algorithm based on k-means is proposed to solve the vehicle-cargo matching problem in this mode. First, according to the geographical location coordinates of the cargo destination, the value of the clustering number k is determined by the silhouette coefficient method, and the cargo clustering scheme is generated using the k-means algorithm. Then, the matching between the cargo clusters and vehicles is completed through a heuristic algorithm based on the fitness value. Then, the path plan for each vehicle is planned. The present invention designs the transfer probability on the basis of the traditional ant colony algorithm, taking into account both time and distance factors, and at the same time adopts the evaporation factor adaptive adjustment strategy and the local search strategy to optimize the path using the improved ant colony algorithm. Whether solving the urban last-mile delivery problem or long-distance transportation, the present invention can obtain reasonable and effective routes. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to the provided drawings.

[0043] Figure 1 It is a schematic flowchart of the overall solution of the present invention;

[0044] Figure 2 It is a flowchart of the K-means clustering algorithm of the present invention;

[0045] Figure 3 It is a schematic diagram of the local search strategy of the present invention;

[0046] Figure 4 It is a flowchart of the multi-strategy improved ant colony path planning algorithm of the present invention;

[0047] Figure 5 It is a schematic diagram of the vehicle-cargo matching problem in the multi-node transfer mode of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0049] The embodiments of the present invention disclose a vehicle-cargo matching method for a multi-node mode, asFigure 1 As shown in the figure, it includes the following steps:

[0050] Determine the number of clusters according to the freight information and vehicle information, and use the k-means algorithm combined with the silhouette coefficient method to cluster the freight information to obtain cargo clusters;

[0051] Match each cargo cluster with the vehicle information based on heuristic rules;

[0052] Use the multi-strategy improved ant colony path planning algorithm to plan the vehicle path and output the optimal solution of vehicle-cargo matching.

[0053] In reality, in the multi-node transfer mode of transportation, often a vehicle carries goods destined for multiple destinations, and these destinations are scattered in different cities. This scattered situation leads to a long transportation path for the goods, increasing the complexity and cost of transportation and reducing the transportation efficiency. Therefore, for this scattered situation of cargo destinations, appropriate methods need to be adopted for destination clustering to improve the efficiency of cargo distribution and reduce costs.

[0054] The present invention selects the k-means clustering algorithm combined with the silhouette coefficient method for clustering. The k-means clustering algorithm is an unsupervised learning algorithm based on distance, which can divide data points into a predetermined number of clusters. Its core idea is to iteratively assign data points to the cluster center closest to them and update the cluster center until convergence.

[0055] Since in the actual matching situation, the number of trucks is often much larger than the amount of goods, and the vehicle types are not the same, the value of k cannot be determined according to the number of trucks, and it is crucial to select the correct number of clustering clusters K. Therefore, the present invention determines the optimal k value through the silhouette coefficient method.

[0056] The silhouette coefficient method evaluates the clustering quality by measuring the ratio of the internal compactness of data points in the cluster to the separation degree between clusters. The steps of the silhouette coefficient method of the present invention are as follows:

[0057] Run K-means clustering: Run each optimal k-means candidate value selected by the elbow method;

[0058] Calculate the silhouette coefficient: For each data point, calculate its silhouette coefficient:

[0059]

[0060] Among them, a represents the average distance from the data point to other points in the same cluster, and b represents the average distance from the data point to the nearest cluster.

[0061] Calculate the average silhouette coefficient: Calculate the average silhouette coefficient corresponding to each optimal k-means candidate value, that is, average the silhouette coefficients of all data points;

[0062] Plot the average silhouette coefficient curve: Plot the average silhouette coefficient for each value of K as a curve graph. Observe the curve. The silhouette coefficient ranges between [-1, 1]. The larger this value, the more reasonable the classification. Thus, determine the value of K (the final number of clusters).

[0063] The heuristic goods allocation method is as follows:

[0064] Determine the quantity of goods in each cluster: Calculate the total weight and total volume of all goods in each cluster, and compare with the vehicle model with the maximum load capacity. If it exceeds, remove the customer farthest from the cluster center until the capacity limit is met;

[0065] Goods reallocation: Check if there is excess vehicle capacity in other customer groups. If so, incorporate the removed customers into the nearest customer group. If not, form a separate cluster;

[0066] Calculate the adaptability δ of each vehicle model: Since variable costs such as vehicle driving costs, fuel costs, and time penalty costs are mainly reflected in the path planning of the next stage, in this stage, the matching effect mainly affects the fixed cost item in the cost. By maximizing the load utilization rate of the vehicle, the expenditure of fixed costs can be most directly reduced, achieving the purpose of cost savings. For each vehicle model, calculate its degree of adaptation to each cluster according to its load capacity. Examine the ratio of the total weight of goods in the cluster to the load of the vehicle. If the ratio is greater than 1, the fitness value is 0 (not considered); if it is less than 1, the closer the fitness value is to 1, the higher the fitness value.

[0067]

[0068] Among them, w i represents the total weight of the i-th cluster, W k represents the load of the k-th vehicle, and δ ik represents the fitness value of the k-th vehicle relative to the i-th cluster.

[0069] Allocate vehicles: Use the greedy algorithm to allocate one by one in the order of adaptability from high to low.

[0070] As Figure 2 shown, the detailed steps of the K-means clustering algorithm of the present invention are as follows:

[0071] S1: Initialization: Set the number of clusters K to be divided, and the initial K cluster centers;

[0072] S2: Allocation: For each data point, i.e., the longitude and latitude coordinates of the destination, calculate its distance from each cluster center, and then allocate the data point to the cluster to which the nearest cluster center belongs. For each data point x i , calculate its distance to each cluster center μ i , and select the nearest cluster center:

[0073]

[0074] where C(x i ) is the cluster to which the data point x i belongs, and μ j is the center of the j-th cluster.

[0075] If the data points are represented as latitude and longitude geographical coordinates, the formula for calculating the distance is:

[0076]

[0077] S3: Update the cluster center: For each cluster, calculate the average value of all data points therein and use it as the new cluster center. The mathematical representation of this step is:

[0078]

[0079] where C j is the number of data points in the cluster C j .

[0080] S4: Repeat S2 and S3 until the cluster center no longer changes significantly, and output the cluster classification result;

[0081] S5: Determine the quantity of goods in each cluster;

[0082] S6: Calculate the adaptability of each type of vehicle;

[0083] S7: Allocate vehicles: Use the greedy algorithm to allocate one by one in the order of adaptability from high to low;

[0084] S8: Update the vehicle cargo status: Once the vehicle allocation is completed, update the status of the cargo in each cluster, record which vehicles are allocated and the quantity of goods transported by each vehicle;

[0085] S9: Iteration: Repeat S5 - 8 until all goods are allocated.

[0086] For the vehicles with goods already allocated, perform path planning for each vehicle to make the path optimal while considering the cargo time window to minimize the total cost.

[0087] The optimization objective of the present invention is to minimize the total cost, including vehicle fixed costs, transportation costs, and time penalty costs. Among them, the transportation costs and time window penalty costs account for the largest proportion in this link. Therefore, in the path search process, these two factors are considered first: waiting time factor and transportation distance factor, so as to minimize the transportation distance during the distribution process and deliver the goods within the specified time range as much as possible. For the above requirements, the present invention uses the transfer probability based on time factor and transportation distance factor to optimize the reasonable selection of path nodes by the genetic algorithm.

[0088] The time factor reflects the urgency of the time requirement for the goods at the goods point. In the present invention, the waiting time factor is defined by the early or late delivery deadline of the goods, so that the goods with an earlier deadline have a higher priority. Suppose there are goods points i and j, where i is the current location and j is the destination. Then the time factor wt can be defined as follows: ij as:

[0089] wt ij = LT i - t j ;

[0090] where: LT j is the latest allowable delivery time of the goods point j; t j is the actual time when the vehicle arrives at point j.

[0091] If the value of wt ij is large, it means that there is enough time margin for the transportation from the goods point i to the goods point j, and there is no hurry for distribution. On the contrary, if the value of wt ij is small or even negative, it means that the deadline of the goods point j is approaching or has even passed, and it is necessary to distribute as soon as possible.

[0092] The transportation distance factor is related to the distance between two goods points. The greater the distance, the greater the transportation distance factor. Suppose d ij represents the distance between customer i and customer j, (x i , y i ) represents the location of customer i, and (x j , y j ) represents the location of customer j. If the transport vehicle departs from customer i and there are two untransported customer points j1 and j2, and the transportation distance factor d ij1 < d ij2 , then the customer j1 with a shorter transportation distance should be preferred for distribution first, which can improve the subsequent transportation efficiency of the vehicle and meet the needs of customers in a timely manner. Therefore, the present invention introduces the transportation distance factor into the transition probability to affect the selection of the transportation path. The mathematical expression of the heuristic information factor d ij is as follows:

[0093] Δy = y2 - y1

[0094] Δx = x2 - x1;

[0095]

[0096] d ij = r·c;

[0097] The average radius r of the earth is taken as 6371 km.

[0098] Since both distance and time in these two parts affect the delivery efficiency and cost, in order to reflect the distance and time factors in the transition probability simultaneously, when designing the transition probability, the present invention constructs a cost function C that comprehensively considers these two factors ij , which simultaneously considers the distance d from the cargo point i to the cargo point j ij and the time factor wt ij , as shown below:

[0099] C ij = χd ij + δwt ij ;

[0100] Wherein, d ij is the distance from the cargo point i to the cargo point j. wt ij is the time factor, which reflects the urgency of the deadline of the cargo point j. χ and δ are weight coefficients, which determine the relative importance of the distance and time factors in the cost function

[0101] In the ant colony algorithm, the design of the visibility n ij (t) depends on the specific nature of the problem and the definition of the cost function. Its meaning is to encourage ants to choose a shorter path. Therefore, the visibility n ij (t) of the present invention is represented by the reciprocal of the cost function:

[0102]

[0103] The transition probability represents the transition probability of the delivery vehicle k from the customer point i to the customer point j. The greater the transition probability of the customer point, the greater the possibility of choosing to deliver this customer. Its calculation formula is as follows:

[0104]

[0105] Assume that after the transport vehicle completes the transport service of the cargo i, and there are still untransported cargo points j1 and j2, calculate the transition probabilities from the cargo point i to the cargo points j1 and j2 respectively and If there is then preferentially select the cargo j2 for transportation

[0106] In the ant colony algorithm, the evaporation of pheromone is a crucial mechanism that mimics the process of pheromone gradually disappearing over time in the real world. The evaporation of pheromone helps prevent the algorithm from prematurely falling into local optimal solutions and drives the algorithm to search for new solutions. Introducing an adaptive pheromone evaporation coefficient is to balance the exploration and exploitation capabilities of the algorithm. The evaporation degree of pheromone can be dynamically adjusted according to the distribution of path lengths during the iteration process. If the path lengths are relatively concentrated, indicating that a relatively good solution has been found, the evaporation can be reduced; conversely, if the path lengths are dispersed, the evaporation is increased to promote the exploration of new paths. The expression formula for the adaptive evaporation coefficient is as follows:

[0107]

[0108] where: L mean is the average length of all paths in the current generation; L min is the shortest length among all paths in the current generation.

[0109] When L mean is close to L min , this indicates that most ants have found similar paths, which may mean that the pheromone distribution points to the neighborhood of an excellent solution. In this case, we hope to retain this valuable pheromone, so the evaporation rate of pheromone is reduced, and thus the value of ρ is smaller.

[0110] On the contrary, when L mean is much greater than L min , it shows that the quality of the paths found by the ants is uneven. At this time, the pheromone distribution may not effectively guide the ants to find good solutions. Therefore, we hope to increase the evaporation rate of pheromone, remove the pheromone that leads to poor paths, and promote the algorithm to explore new paths. Thus, the value of ρ is larger.

[0111] In this way, the adaptive pheromone evaporation coefficient can dynamically adjust the evaporation degree of pheromone according to the current path length distribution, which helps to maintain the balance between the algorithm's exploration of new solutions and the exploitation of known excellent solutions.

[0112] In the ant colony algorithm, local search methods are used to improve the quality of solutions by making small adjustments to the current solution to find better ones. The 2-opt operator is an optimization algorithm for solving the vehicle routing problem. Its basic idea is to randomly select two nodes in the path and reverse the path between these two nodes to eliminate possible crossings in the path and achieve path optimization. If the reversed path is shorter than the original path, the reversed path replaces the original one. Suppose there is a path {0, 2, 4, 3, 1, 0}, and the 2-opt operator selects nodes 2 and 1 for operation, reversing the path between these two nodes, the new path obtained is {0, 2, 3, 4, 1, 0}. According to the triangle inequality, if the length of the new path is shorter than that of the original path, the new path is accepted. This is because in geometric space, the length of any side of a triangle formed by any three points is less than the sum of the lengths of the other two sides. Therefore, if there is a crossing formed by nodes 1, 2, 3, and 4 in the original path, then by reversing this part of the path, this crossing can be eliminated, thus possibly shortening the total length of the path, as Figure 3 shown.

[0113] As Figure 4 shown, the overall process of the improved strategy ant colony algorithm is as follows:

[0114] Initialize parameters: Initialize the number of ants m, and randomly place all ants on n different cities. Set time t = 0, iteration number iter = 0, and set the maximum iteration number iter max . Initialize the pheromone matrix Δτ ij (0) = 0, initialize the visibility matrix n ij (t), and set the constant ρ0 as the initial pheromone concentration. Set the information heuristic factor α and the expected heuristic factor β. Set the pheromone increment intensity Q. Set the taboo list to record the cities visited by the ants;

[0115] Generate paths: Based on the transition probability calculate the jump probability of accessible nodes, use the roulette wheel strategy to select the jump nodes of the ants, and update the taboo list tabulist k , and add the selected cities to it. Repeat the operation until a complete path is constructed, that is, all cities are visited and returned to the starting city;

[0116] Calculate the fitness value: Calculate the cost of each path, and record the optimal cost and corresponding path of the current generation;

[0117] Update pheromone: Calculate the adaptive evaporation coefficient and update the pheromone concentration τ ij (t + 1);

[0118] Determine whether the algorithm ends: If the maximum number of iterations iter is reached max , the algorithm stops running and outputs the optimal result. If the maximum number of iterations iter is not reached max , clear the taboo list of all ants and return to "generate path" to continue the iteration;

[0119] Output result: Output the best path found and its corresponding objective function value.

[0120] In the field of road freight, a multi-node transfer network refers to a logistics network that involves multiple transfer points or transfer stations during the process of goods transportation. In such a network, goods start from the origin, may pass through one or more transfer stations, and finally reach the destination. For example, less-than-truckload freight involves combining small batches of goods destined for different destinations for transportation and needs to be delivered to several different cities in sequence; large logistics companies usually have a transportation network covering a wide area, and goods may need to be sorted, reloaded, and transferred at several transfer stations before reaching the final destination; in the last-mile transportation, goods may first be transported from a distribution center to nearby regional outlets. Under the multi-node transfer network, the vehicle-cargo matching problem is particularly important, and it is necessary to comprehensively consider the allocation of goods and the scheduling of vehicles to optimize the efficiency and cost of the entire transportation process. The problem is illustrated as Figure 5 shown.

[0121] To fit the actual application scenario and simplify the model to a certain extent, the following assumptions are made first:

[0122] There is a batch of goods with different volumes, weights, destinations, and timeliness;

[0123] There is a batch of available vehicles with different models, load capacities, and volumes;

[0124] These goods need to be matched with vehicles, while ensuring that the maximum load and volume limits of each vehicle are not exceeded, and the vehicles transport the allocated goods in sequence;

[0125] The goal is to find a matching plan for goods and vehicles so that all goods can be transported to the corresponding destinations, while minimizing the total transportation cost;

[0126] Without considering the special attributes of goods, the goods from different sources can be assembled;

[0127] This embodiment is based on the study of the matching of local vehicle sources and goods sources in the city. Therefore, it is defaulted that the starting point of the goods and the location of the vehicles are the same;

[0128] In this embodiment, the research scenario is that all goods are pre-transported to the distribution point in the city by small vans before transportation. Freight vehicles directly load the goods from this point and depart, without having to pick up goods at multiple locations. Therefore, there is no pick-up link in this model;

[0129] Each good corresponds to a destination, which is called a good node;

[0130] The vehicle departs from the shipping point and finally needs to return to the shipping point to complete the transportation;

[0131] It is assumed that idle vehicles can meet the needs of all shippers;

[0132] If the vehicle arrives later than the given deadline of the good, a corresponding time penalty cost will be incurred, but no penalty cost will be incurred if it arrives earlier than the specified time;

[0133] Goods cannot be split, and each good can only be transported by one vehicle;

[0134] The vehicle travels at a constant speed and generates a corresponding service time at each good node.

[0135] In this embodiment, the objective function mainly includes three parts, namely vehicle fixed cost, transportation cost, and time penalty cost.

[0136] Fixed cost analysis: Enabling a vehicle will incur a certain fixed cost, which increases with the increase in the number of enabled vehicles and is independent of the vehicle's driving mileage and the number of customers visited. When the number of enabled vehicles increases or decreases, the total cost also changes accordingly. If the fixed cost of vehicle k is set as f k , then the expression of the fixed cost TC1 is as follows:

[0137]

[0138] Transportation cost analysis: In multi-node transportation, a freight vehicle usually needs to visit multiple good nodes. When vehicle k visits the path (i, j), the driving distance is d ijk . Therefore, the expression of the transportation cost TC2 is as follows:

[0139]

[0140] Time penalty cost analysis: This embodiment mainly considers the soft time window scenario. If the transportation arrival time is later than the customer-given time, a corresponding penalty cost will be incurred. When the vehicle arrives later than the deadline, the late penalty cost is included. The expression of the time window penalty cost function p(t) is as follows:

[0141]

[0142] Then the time window deviation penalty cost TC3 is:

[0143]

[0144] In summary, the objective function and the expression of the total cost in this embodiment are as follows:

[0145] minf(x) = FC + VC + TC;

[0146] The constraint conditions are as follows:

[0147] Vehicle scheduling constraint: The vehicle route is a closed loop, and both the starting point and the ending point are the shipping points.

[0148]

[0149] Secondly, each customer is served by exactly one vehicle during the transportation process.

[0150]

[0151] Each node should maintain the balance of input and output to ensure that the number of times a vehicle travels from node i to node j is equal to the number of times the vehicle travels from node j to node i. This balance ensures that after completing the task at each node, the vehicle will continue to the next node, thus maintaining the continuity of the path. This helps to ensure that the transport vehicle can execute the task efficiently and avoid unnecessary stops and path interruptions.

[0152]

[0153] Each vehicle can be dispatched at most once for each match.

[0154]

[0155] Travel time constraint: The time for the vehicle to depart from the transportation center and reach customer i is the time of departure from the shipping point plus the travel time of the path (0, j). The time for the vehicle to travel from i to j is the time of arrival at i plus the service time and the time on the way (i, j).

[0156]

[0157] The driving process of the vehicle cannot exceed the maximum driving time T of the vehicle k .

[0158]

[0159] Vehicle load constraint:

[0160] The load of the vehicle at any time shall not exceed the maximum load and volume limit of the vehicle.

[0161]

[0162] When the vehicle accesses the path (i, j) and arrives at customer point j, the load capacity of the vehicle is equal to the load capacity of the vehicle when it arrives at customer point i minus the required weight of customer point i.

[0163]

[0164] Finally, the present invention also discloses a vehicle-cargo matching system for a multi-node mode, including:

[0165] Cargo clustering module: used to determine the number of clusters according to freight information and vehicle information, and use the k-means algorithm combined with the silhouette coefficient method to cluster the freight information to obtain cargo clusters;

[0166] Vehicle-cargo matching module: used to match each cargo cluster with vehicle information based on heuristic rules;

[0167] Optimal path planning module: used to plan the vehicle path using a multi-strategy improved ant colony path planning algorithm and output the optimal solution of vehicle-cargo matching.

[0168] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0169] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined in the present invention can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown in the present invention, but will conform to the widest scope consistent with the principles and novel features disclosed in the present invention.

Claims

1. A vehicle-cargo matching method for a multi-node mode, characterized in that: The following steps are involved: The number of clusters is determined based on the freight information and vehicle information, and the freight information is clustered using the k-means algorithm combined with the silhouette coefficient method to obtain the freight clusters; Match each cargo cluster with vehicle information based on heuristic rules; Use multiple strategies to improve the ant colony path planning algorithm to plan vehicle paths and output the optimal solution for vehicle-cargo matching; Based on the heuristic rules, each cargo cluster is matched with the vehicle information. The specific steps are as follows: Determine the cargo volume of each cluster: Calculate the total weight and volume of all cargo in each cluster and compare it with the model with the maximum load capacity. If it exceeds, remove the customer farthest from the cluster center until the capacity limit is met; Cargo redistribution: Determine whether there is excess vehicle capacity in other customer groups. If so, the excluded customers will be included in the nearest customer group. If not, they will become a separate cluster. Calculate the adaptability of each vehicle model using the following formula: Among them, w i represents the total weight of the i-th cluster, W k represents the load of the kth vehicle, δ ik It represents the fitness value of the k-th vehicle relative to the i-th cluster; A greedy algorithm is used to assign vehicles one by one in the order of their adaptability; The specific steps of the multi-strategy improved ant colony path planning algorithm are as follows: Initialize parameters and set the maximum number of iterations iter max ; Based on the transition probability Calculate the jump probability of accessible nodes, use the roulette strategy to select the jump node of the ant, and update the tabulist k , add the selected city to it, and repeat the operation until a complete path construction is completed, that is, all cities are visited and returned to the starting city; Calculate the cost of each road path and record the current optimal cost and corresponding path; Calculate the adaptive evaporation coefficient and update the pheromone concentration τ on each path ij (t+1); If the maximum number of iterations iter is reached max , the algorithm stops running and outputs the optimal result; if the maximum number of iterations iter is not reached max , then clear the taboo tables of all ants and continue iteration; Output the best path found and its corresponding objective function value; Transition probability The calculation formula is as follows: Among them, n ij (t) represents the visibility in the heuristic, α represents the time penalty coefficient, β represents the expected heuristic factor, τ ij (t) represents pheromone; Visibility in inspiration ij The design of (t) depends on the specific nature of the problem and the definition of the cost function. The meaning is to encourage ants to choose paths with shorter distances. The visibility n in the inspiration ij (t) is expressed as the inverse of the cost function: Cost function C ij , while considering the distance d from cargo point i to cargo point j ij and time factor wt ij , as shown below: C ij =χd ij +δwt ij ; Among them, d ij is the distance from cargo point i to cargo point j, wt ij is the time factor, reflecting the urgency of the deadline of cargo point j, and χ and δ are weight coefficients; The expression formula of adaptive evaporation coefficient is as follows: Among them, L mean represents the average length of all contemporary paths; L min Indicates the shortest length of all contemporary paths.

2. A vehicle-cargo matching method for a multi-node mode according to claim 1, characterized in that: The steps to determine the cluster value k using the silhouette coefficient method are as follows: Run K-means clustering: Run each best k-means candidate selected by the elbow method; Calculate the silhouette coefficient: For each data point, calculate its silhouette coefficient: Among them, a represents the average distance from the data point to other points in the same cluster, and b represents the average distance from the data point to the nearest cluster; Calculate the average silhouette coefficient: Calculate the average silhouette coefficient corresponding to each best k-means candidate value, that is, average the silhouette coefficients of all data points; Draw the average silhouette coefficient curve: Draw the average silhouette coefficient of each k value into a curve graph; observe the curve, the silhouette coefficient range is between [-1,1], the larger the value, the more reasonable the classification, thus determining the K value.

3. A vehicle-cargo matching system for a multi-node mode, using a vehicle-cargo matching method for a multi-node mode according to any one of claims 1-2, characterized in that: include: Cargo clustering module: used to determine the number of clusters based on freight information and vehicle information, and cluster freight information using the k-means algorithm combined with the silhouette coefficient method to obtain cargo clusters; Vehicle-cargo matching module: used to match each cargo cluster with vehicle information based on heuristic rules; Optimal path planning module: used to improve the ant colony path planning algorithm using multiple strategies to plan vehicle paths and output the optimal solution for vehicle-cargo matching.

Images