A location-route decision method and related device for joint delivery of trucks and unmanned vehicles

Abstract

The present invention provides a site selection and path decision method and related devices for the joint delivery of trucks and unmanned vehicles, and relates to the field of delivery planning technology. When a delivery task is received, candidate distribution points for unmanned vehicles and their service ranges are obtained according to the delivery task; customer needs are obtained according to the delivery task, and the distribution point selection threshold is dynamically adjusted to determine whether each candidate distribution point is enabled; for the enabled distribution points, the joint delivery path of trucks and unmanned vehicles is optimized through an improved adaptive large neighborhood search algorithm to minimize the total delivery cost; the total delivery cost, the optimal path, and the distribution point selection status are calculated and output. For the delivery mode of joint delivery of trucks and unmanned vehicles, a dynamic site selection strategy for unmanned vehicle distribution points based on delivery demand is proposed to flexibly allocate delivery tasks, reduce manual workload, and improve delivery efficiency.

Description

Technical Field

[0001] The present application relates to the technical field of distribution planning, and in particular to a site selection and path decision-making method and related devices for joint distribution by trucks and unmanned vehicles. Background Art

[0002] In recent years, with the rapid development of e-commerce, demand for logistics and distribution has exploded. Within the entire logistics and distribution process, last-mile delivery is considered the most challenging link and the one with the lowest consumer satisfaction. The choice of delivery strategy directly impacts logistics efficiency, costs, and customer satisfaction. Traditional last-mile delivery models rely on manual, repetitive delivery tasks, resulting in not only high workloads but also high operating costs due to the massive labor investment. Coupled with the increasingly severe problem of an aging population, a declining labor force, and rising labor costs, traditional delivery models are increasingly unable to meet the growing demand for last-mile delivery.

[0003] Against this backdrop, intelligent delivery technologies, such as drones and unmanned vehicles, offer new solutions for revolutionizing end-to-end logistics. Unmanned delivery systems utilize automation and intelligent equipment to reduce reliance on human labor for delivery, enabling 24 / 7 continuous operation. These systems significantly improve delivery efficiency and reduce labor costs, demonstrating broad application prospects. Current research focuses on collaborative delivery models using trucks and drones, but this approach still faces limitations in high-density scenarios such as urban communities. For example, drones, limited by their payload capacity, require multiple round trips to reach high-demand areas, resulting in longer flight distances and increased energy consumption and operating costs. Furthermore, extreme weather conditions can impact their stability and adaptability. In contrast, unmanned vehicles offer significant advantages in payload capacity, range, and environmental adaptability, making them more suitable for large-scale delivery in high-demand areas.

[0004] Therefore, how to flexibly allocate delivery tasks, reduce manual workload, and improve delivery efficiency has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The core of the present invention is to provide a site selection and route decision-making method and related devices for joint delivery of trucks and unmanned vehicles, which solves the problems of flexibly allocating delivery tasks, reducing manual workload and improving delivery efficiency.

[0006] In the first aspect, the present application provides a method for site selection and routing decision-making for joint delivery by trucks and unmanned vehicles, which adopts the following technical solutions:

[0007] A location and routing decision method for combined delivery of trucks and unmanned vehicles, comprising:

[0008] Upon receiving a delivery task, obtaining candidate collection and distribution points for unmanned vehicle delivery and their service ranges according to the delivery task;

[0009] Obtain customer demand based on the delivery task, and dynamically adjust the distribution point selection threshold to determine whether each candidate distribution point is enabled;

[0010] For enabled distribution points, an improved adaptive large neighborhood search algorithm is used to optimize the joint delivery routes of trucks and unmanned vehicles to minimize the total delivery cost.

[0011] Calculate and output the total delivery cost, optimal route, and distribution point selection status, where the total delivery cost includes the truck driving cost, the unmanned vehicle usage time cost, the driver labor cost, and the fixed usage cost of the distribution point.

[0012] Optionally, the dynamically adjusting the collection and distribution point selection threshold specifically includes:

[0013] Calculate the difference between the delivery cost of unmanned vehicles when the distribution point is enabled and the delivery cost of trucks when it is not enabled;

[0014] Dynamically update the selection threshold based on the difference. If the delivery cost of the unmanned vehicle is lower, then increase the selection threshold to increase the probability of the distribution point being activated. Otherwise, decrease the selection threshold.

[0015] The normal distribution function is used to determine whether the collection and distribution point is enabled, and a collection and distribution point selection state coding sequence is generated.

[0016] Optionally, the improved adaptive large neighborhood search algorithm includes:

[0017] The destruction operators are used to perform local destruction on the truck routes, including random destruction, worst cost destruction and correlated destruction.

[0018] Use repair operators to reconstruct the damaged path, including random repair, greedy repair and regret repair;

[0019] Design neighborhood search operators for autonomous vehicle paths, including Swap, Insert, 2-opt, and or-opt operators;

[0020] The simulated annealing acceptance criterion is introduced to accept poor solutions in a probabilistic manner to prevent the algorithm from falling into local optimality.

[0021] Optionally, the similarity of the destruction operator is calculated as:

[0022]

[0023] in, and is the weight coefficient, Points for customers To the customer point distance, and Customer points and customer points If the customer orders and customer points On the same path If it is 0, otherwise it is 1, and the similarities are removed from large to small until the specified number of destructions is reached.

[0024] Optionally, the step of reconstructing the damaged path using a repair operator includes:

[0025] Obtain the operator weight and select one of the three repair operators according to the roulette wheel method to repair the truck path and obtain the solution. ;

[0026] Among them, random repair includes: randomly selecting a customer point from the set of damaged customer points and inserting it into a random position of the truck delivery path until there are no remaining customer points to be inserted;

[0027] Greedy repair includes: selecting a customer point from the set of destroyed customer points, calculating the incremental delivery cost when the customer point is inserted into different locations of different delivery routes of the truck, and selecting the insertion position with the smallest incremental delivery cost and inserting it into that position until there are no remaining customer points to be inserted.

[0028] Regret repair includes: calculating the delivery cost of each customer in the damaged customer point set at the optimal insertion position and the delivery cost of the suboptimal insertion position, taking the difference between the costs of the two positions as the regret value, and selecting the customer point with the largest regret value and inserting it into the optimal position.

[0029] Optionally, the step of designing a neighborhood search operator for the unmanned vehicle path includes:

[0030] According to the roulette method, one of the four neighborhood search operators is selected to obtain the delivery unmanned vehicle path. ;

[0031] Among them, the Swap operator exchanges the positions of two random customer points in the unmanned vehicle delivery path;

[0032] The Insert operator randomly selects two customer points on the delivery path of the unmanned vehicle. , the customer points Insert Customer Point after;

[0033] The 2-opt operator randomly selects two client nodes , the client node With the client node The customer positions between them are in reverse order;

[0034] The or-opt operator randomly selects two customer points on the unmanned vehicle delivery path. , the customer points and customer points Flip and insert into the customer point after.

[0035] Optionally, the mathematical model of the total delivery cost is:

[0036]

[0037] in, is the distance traveled by the truck, The time cost of using the autonomous vehicle, Labor costs for drivers, The usage cost of the distribution point is fixed, and the path flow conservation and capacity constraints are met.

[0038] In a second aspect, the present application provides a location selection and routing decision-making device for combined delivery of trucks and unmanned vehicles, which executes the method described above, including:

[0039] A data acquisition module is used to obtain candidate collection and distribution points for unmanned vehicle delivery and their service ranges based on the delivery task when a delivery task is received;

[0040] A judgment module is used to obtain customer demand based on the delivery task and dynamically adjust the collection and distribution point selection threshold to determine whether each candidate collection and distribution point is enabled;

[0041] The algorithm module optimizes the joint delivery routes of trucks and unmanned vehicles for enabled distribution points through an improved adaptive large neighborhood search algorithm to minimize the total delivery cost.

[0042] The calculation module is used to calculate and output the total distribution cost, the optimal route and the distribution point selection status, wherein the total distribution cost includes the truck driving cost, the unmanned vehicle usage time cost, the driver labor cost and the fixed usage cost of the distribution point.

[0043] In a third aspect, the present application provides a computer device, comprising: a memory and a processor, wherein the processor executes the method described above when running computer instructions stored in the memory.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium comprising instructions, which, when executed on a computer, enable the computer to execute the method described above.

[0045] In summary, this application has the following beneficial technical effects:

[0046] This application obtains candidate collection and distribution points for unmanned vehicles and their service areas based on the delivery task upon receiving the delivery task; dynamically adjusts the collection and distribution point selection threshold to determine whether each candidate collection and distribution point is enabled based on the delivery task; optimizes the combined delivery path of trucks and unmanned vehicles using an improved adaptive large neighborhood search algorithm for enabled collection and distribution points to minimize the total delivery cost; and calculates and outputs the total delivery cost, optimal path, and collection and distribution point selection status. For the delivery model of combined truck and unmanned vehicle delivery, a dynamic collection and distribution point selection strategy for unmanned vehicles based on delivery demand is proposed to flexibly allocate delivery tasks, reduce manual workload, and improve delivery efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the computer device structure of the hardware operating environment involved in the embodiment of the present application.

[0048] Figure 2 This is a flow chart of the first embodiment of the site selection and route decision-making method for joint delivery by trucks and unmanned vehicles in the present application.

[0049] Figure 3 This is a conceptual diagram of the site selection and route planning process for the joint delivery of trucks and unmanned vehicles in this application.

[0050] Figure 4 This is a conceptual diagram of the site selection strategy for this application.

[0051] Figure 5 This is the flowchart of the improved adaptive large neighborhood search algorithm of this application.

[0052] Figure 6 is a schematic diagram of the neighborhood operator, Figure 6 (a) is a schematic diagram of swap, Figure 6 (b) is the Insert diagram. Figure 6 (c) is a schematic diagram of 2-opt, Figure 6 (d) is the schematic diagram of or-opt.

[0053] Figure 7 This is the location selection and path planning diagram of the unmanned vehicle for cases C101, C201, R101, and RC202. Figure 7 (a) is the C101 site selection path planning map, Figure 7 (b) is the C201 site selection path planning map, Figure 7 (c) is the R101 site selection path planning map, Figure 7 (d) is the RC101 site selection path planning map.

[0054] Figure 8 This is a structural block diagram of the first embodiment of the site selection and path decision-making device for joint delivery by trucks and unmanned vehicles in the present application. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below through the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0056] Reference Figure 1 , Figure 1 This is a schematic diagram of the computer device structure of the hardware operating environment involved in the embodiment of the present application.

[0057] like Figure 1 As shown, the computer device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display and an input unit, such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a wireless fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also be a storage device independent of the processor 1001.

[0058] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the computer device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0059] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a location selection and routing decision program for the joint delivery of trucks and unmanned vehicles.

[0060] exist Figure 1In the computer device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in this application can be set in the computer device, and the computer device calls the site selection-path decision program for the joint delivery of trucks and unmanned vehicles stored in the memory 1005 through the processor 1001, and executes the site selection-path decision method for the joint delivery of trucks and unmanned vehicles provided in the embodiment of this application.

[0061] The embodiment of the present application provides a method for site selection and route decision making for joint delivery of trucks and unmanned vehicles, referring to Figure 2 , Figure 2 This is a flow chart of the first embodiment of the site selection and route decision-making method for joint delivery by trucks and unmanned vehicles in this application.

[0062] In this embodiment, the location selection and routing decision method for combined delivery of trucks and unmanned vehicles includes the following steps:

[0063] Step S10: When a delivery task is received, candidate collection and distribution points for unmanned vehicle delivery and their service ranges are obtained according to the delivery task.

[0064] In specific implementation, the explanation of terms in this embodiment includes:

[0065] VRP: Vehicle Routing Problem, a classic optimization problem in operations research, aims to design the best route for a fleet of vehicles to serve a group of customers at the lowest cost (such as distance and time) while satisfying constraints (such as vehicle capacity and time window).

[0066] Heuristic algorithms are algorithms that use empirical or intuitive rules to quickly find feasible solutions to problems. Their core goal is to find a "good enough" solution within a reasonable time, rather than guaranteeing a mathematically optimal solution. They are particularly suitable for complex optimization problems (such as NP-hard problems), where traditional exact algorithms (such as exhaustive enumeration and dynamic programming) are computationally prohibitive.

[0067] Distribution point: A station used for trucks to unload goods, unmanned vehicles to dock, and unmanned vehicles to recharge, similar to a transfer station.

[0068] It should be noted that the technical problems in the existing scenario in this embodiment include:

[0069] (1) In existing studies, most of the methods used were single-vehicle delivery or coordinated delivery by trucks and drones. The former uses manual driving to complete the entire process, which faces greater delivery pressure in areas with high demand. The latter, through the division of labor and cooperation between trucks and drones, has alleviated the pressure of manual delivery to a certain extent. However, due to the limited load capacity of drones, delivery tasks need to be carried out frequently in high-density demand areas, resulting in an increase in the frequency of equipment charging, which ultimately reduces the overall delivery efficiency.

[0070] (2) Due to the large size of unmanned vehicles, they need to be stored and recharged at distribution points. In existing scenarios, for transportation problems involving distribution points, trucks are usually set to only undertake trunk transportation between distribution points, and customer delivery tasks are entirely handled by short-distance vehicles equipped at distribution points. This distribution model fails to fully consider the delivery capabilities of trucks along the way when allocating tasks. When the customer point is in the middle area between two distribution points, if delivery is only carried out by vehicles at the distribution points, the vehicles will inevitably have to make long-distance round trips, which not only increases the delivery distance but also limits the space for path optimization.

[0071] (3) Current research generally adopts a fixed distribution point location model, that is, once the distribution point location is determined, it remains in operation for a long time. Research on dynamic distribution point location selection that responds to dynamic changes in customer demand is relatively scarce. In actual logistics scenarios, customer distribution is complex and changeable, and customer distribution has a significant impact on delivery efficiency and cost. An operation model that relies solely on fixed distribution points is difficult to adapt to the complex and changing customer distribution characteristics, resulting in insufficient operational flexibility.

[0072] Therefore, this embodiment aims to address the numerous and repetitive round-trip delivery tasks required within a certain range. For example, in urban communities with large populations, the daily delivery tasks are numerous. Relying solely on manual labor creates significant delivery pressure, while drones, limited by their payload capacity, cannot fully leverage their advantages in such demand-intensive areas. Furthermore, the existing distribution model with hubs and hubs is inflexible in allocating delivery tasks to address the complex and changing customer distribution landscape, limiting the scope for route optimization.

[0073] In practice, to address issues such as excessive manual repetitiveness and low delivery efficiency in terminal delivery, this embodiment employs a truck-autonomous vehicle (AV) combined delivery model to serve customers. The logistics system consists of a distribution center, candidate AV delivery collection points, customers, trucks, and AVs. Each customer's coordinates, demand, required service time, truck capacity, AV capacity, coordinates of candidate AV delivery collection points, and service areas are known. Multiple AV delivery collection points are pre-set based on factors such as customer distribution and demand. Each collection point has multiple AVs for delivery. AVs can only deliver to customers within the service area of ​​their collection point. Customers outside of the collection point's service area are delivered by trucks. During logistics delivery, a suitable AV collection point is first selected based on the AV collection point dynamic location selection strategy. Based on this, trucks load goods and depart from the distribution center to deliver to customers. Upon arrival at the AV collection point, trucks unload goods for all customers within the collection point's service area, and then AVs deliver to customers in that area. If a candidate AV collection point is not selected, trucks deliver to all customers within the collection point's service area. After completing the delivery task, the truck returns to the distribution center, and the unmanned vehicle returns to the original distribution point after completing the delivery task. The specific process is as follows Figure 3 shown.

[0074] It should be noted that the specific applicable scope in this embodiment corresponds to the following conditions:

[0075] 1. The trucks and unmanned vehicles are of the same type;

[0076] 2. The demand from each customer within the autonomous vehicle distribution point area is less than the autonomous vehicle capacity;

[0077] 3. Each customer can be served only once by a self-driving car or truck;

[0078] 4. Each unmanned vehicle can be used for delivery at most once, and each truck can be used for delivery at most once.

[0079] In specific implementation, the total delivery cost includes the truck driving cost, the truck driver's salary cost, the truck's time cost of serving customers and the time cost of serving distribution points, the unmanned vehicle's time cost, and the fixed usage cost of distribution points.

[0080] Step S20: Obtain customer demand based on the delivery task, and dynamically adjust the collection and distribution point selection threshold to determine whether each candidate collection and distribution point is enabled.

[0081] It should be noted that the dynamic adjustment of the distribution point selection threshold specifically includes: calculating the difference between the delivery cost of the unmanned vehicle when the distribution point is enabled and the delivery cost of the truck when it is not enabled; dynamically updating the selection threshold based on the difference. If the delivery cost of the unmanned vehicle is lower, the selection threshold is increased to increase the probability of enabling the distribution point, otherwise the selection threshold is lowered; judging whether the distribution point is enabled through the normal distribution function, and generating a distribution point selection status coding sequence.

[0082] In specific implementation, the dynamic location selection strategy for unmanned vehicle distribution points includes:

[0083] The site selection of unmanned vehicle distribution points is a dynamic and cyclical optimization process. The selection threshold is initially set to 0.5, and the goal is to deliver to all customers in the enabled distribution point area at the lowest cost.

[0084] Specific site selection strategy steps are as follows: Figure 4 shown.

[0085] exist Figure 4 The corresponding site selection process includes:

[0086] Step 1: Calculate the delivery cost of the distribution point area. In the current iteration, the unmanned vehicle path in the distribution point area is optimized. Based on the optimization results, the cost is calculated in the following two ways: ① Calculate the cost of the unmanned vehicle from the distribution point The cost of returning to the original distribution point after completing the delivery task for all customer points in the area, the fixed cost of using the distribution point, and the cost of trucks entering and leaving the distribution point from the customer point closest to the distribution point area. The sum of the costs (the distribution cost when the corresponding distribution point is selected); ② Calculate the distribution point for direct delivery by truck The sum of the delivery costs of all customer points within the distribution point area and the customer points closest to the distribution point area when driving into and out of the distribution point area (the delivery cost when the corresponding distribution point is not selected).

[0087] Step 2: Update the selection threshold. By comparing the two costs in step 1, dynamically adjust the collection and distribution points according to the following formula The selection threshold.

[0088]

[0089] in, The collection and distribution points in this iteration The selection threshold of The scatter points in the previous iteration The selection threshold of Trucks at the distribution point delivery costs, For unmanned vehicles at the distribution point delivery costs, For trucks and unmanned vehicles at the distribution point The difference in delivery costs. is the attenuation coefficient.

[0090] Step 3: Determine the selection status of the collection and distribution points. The selection threshold is used to determine whether the scatter point is selected in the next iteration. If the threshold condition is met (that is, the function value of the normal distribution between [0,1] is less than the selection threshold ), then the gathering and distribution points The code is 1 (indicating that the collection and distribution point is selected), otherwise, the code is 0 (indicating that the collection and distribution point is not selected).

[0091] Throughout the site selection optimization process, if the cost of using autonomous vehicles for delivery within a particular distribution point is calculated to be lower than that of truck delivery, the selection threshold for that distribution point will be raised, thereby increasing the probability of it being selected in the next iteration. Conversely, if the cost of truck delivery is calculated to be lower, the selection threshold for that distribution point will be lowered, thereby decreasing its probability of selection. As the optimization process progresses, the selection status of distribution points gradually converges, and distribution points with lower autonomous vehicle delivery costs when selected will be prioritized in subsequent iterations.

[0092] Step S30: For the enabled distribution points, the improved adaptive large neighborhood search algorithm is used to optimize the joint delivery path of trucks and unmanned vehicles to minimize the total delivery cost.

[0093] In a specific implementation, the improved adaptive large neighborhood search algorithm includes:

[0094] The destruction operators are used to perform local destruction on the truck routes, including random destruction, worst cost destruction and correlated destruction.

[0095] Use repair operators to reconstruct the damaged path, including random repair, greedy repair and regret repair;

[0096] Design neighborhood search operators for autonomous vehicle paths, including Swap, Insert, 2-opt, and or-opt operators;

[0097] The simulated annealing acceptance criterion is introduced to accept poor solutions in a probabilistic manner to prevent the algorithm from falling into local optimality.

[0098] It should be noted that the similarity of the destruction operator is calculated as:

[0099]

[0100] in, and is the weight coefficient, Points for customers To the customer point distance, and Customer points and customer points If the customer orders and customer points On the same path If it is 0, otherwise it is 1, and the similarities are removed from large to small until the specified number of destructions is reached.

[0101] It is understandable that the step of reconstructing the damaged path using the repair operator includes:

[0102] Obtain the operator weight and select one of the three repair operators according to the roulette wheel method to repair the truck path and obtain the solution. ;

[0103] Among them, random repair includes: randomly selecting a customer point from the set of damaged customer points and inserting it into a random position of the truck delivery path until there are no remaining customer points to be inserted;

[0104] Greedy repair includes: selecting a customer point from the set of destroyed customer points, calculating the incremental delivery cost when the customer point is inserted into different locations of different delivery routes of the truck, and selecting the insertion position with the smallest incremental delivery cost and inserting it into that position until there are no remaining customer points to be inserted.

[0105] Regret repair includes: calculating the delivery cost of each customer in the damaged customer point set at the optimal insertion position and the delivery cost of the suboptimal insertion position, taking the difference between the costs of the two positions as the regret value, and selecting the customer point with the largest regret value and inserting it into the optimal position.

[0106] It should be noted that the steps of designing a neighborhood search operator for the unmanned vehicle path include:

[0107] According to the roulette method, one of the four neighborhood search operators is selected to obtain the delivery unmanned vehicle path. ;

[0108] Among them, the Swap operator exchanges the positions of two random customer points in the unmanned vehicle delivery path;

[0109] The Insert operator randomly selects two customer points on the delivery path of the unmanned vehicle. , the customer points Insert Customer Point after;

[0110] The 2-opt operator randomly selects two client nodes , the client node With the client node The customer positions between them are in reverse order;

[0111] The or-opt operator randomly selects two customer points on the unmanned vehicle delivery path. , the customer points and customer points Flip and insert into the customer point after.

[0112] Step S40: Calculate and output the total delivery cost, optimal route and distribution point selection status, wherein the total delivery cost includes the truck driving cost, the unmanned vehicle usage time cost, the driver labor cost and the distribution point fixed usage cost.

[0113] In a specific implementation, the mathematical model of the total delivery cost is:

[0114]

[0115] in, is the distance traveled by the truck, The time cost of using the autonomous vehicle, Labor costs for drivers, The usage cost of the distribution point is fixed, and the path flow conservation and capacity constraints are met.

[0116] In a specific implementation, the mathematical model of this embodiment is:

[0117] (1) Collection. A collection of candidate distribution points for unmanned vehicle delivery; Candidate distribution points for unmanned vehicle delivery The collection of all customers within the service scope; The customer collection for all candidate collection and distribution points for unmanned vehicle delivery. ; Collect all customers outside the candidate collection and distribution point area for unmanned vehicle delivery; For all customers, ; is the set of all nodes in the logistics system, ,in For distribution centers; muster for the wagons; Gather for driverless cars; For all vehicles, .

[0118] (2)Symbol. is the truck capacity; The capacity for unmanned vehicles; For customers Demand; Distribution points for unmanned vehicles The sum of the demands of all customers within the service scope; For nodes To Node distance; For vehicles Arrival Node moment; For vehicles Slave nodes the moment of departure; For customers Required service time; For distribution points Required service time; For driverless cars Arrival at customer point The remaining driving distance when For driverless cars Leave the node The remaining driving distance when For vehicles Arrival Node Real-time load at the time; For vehicles Leave the node Real-time load at the time; is the speed of the truck; is the driving speed of the autonomous vehicle; For trucks Slave nodes To Node driving time; For driverless cars Slave nodes To Node driving time.

[0119] (3) Decision variables. Is a 0-1 variable, if the vehicle Access Node , the value is 1, otherwise the value is 0; is a 0-1 variable. If the alternative unmanned vehicle distribution point If enabled, the value is 1, otherwise it is 0; Is a 0-1 variable, if the vehicle Slave nodes Travel to the node , the value is 1, otherwise the value is 0;

[0120] It can be understood that, in this embodiment, the distribution cost analysis includes:

[0121] make is the cost per unit distance traveled by trucks (unit: yuan / km), and the cost per unit distance traveled by all trucks is for

[0122] (1)

[0123] make The unit time cost of the unmanned vehicle (unit: yuan / minute), the total time cost of all unmanned vehicles for

[0124] (2)

[0125] make The cost per unit time of the driver (unit: yuan / minute). The labor cost includes the time cost of the truck driving, the time cost of the truck delivering to customers and distribution points, and the labor cost of all drivers. for

[0126] (3)

[0127] make The fixed cost of using a candidate distribution point for unmanned vehicle delivery when it is enabled for delivery (unit: yuan / point), and the cost of using all enabled distribution points for

[0128] (4)

[0129] 3 Mathematical model

[0130] With the goal of minimizing total delivery cost, the location selection and routing planning model for joint delivery of trucks and unmanned vehicles is constructed as follows:

[0131] (5)

[0132] (6)

[0133] (7)

[0134] (8)

[0135] (9)

[0136] (10)

[0137] (11)

[0138] (12)

[0139] (13)

[0140] (14)

[0141] (15)

[0142] (16)

[0143] (17)

[0144] (18)

[0145] (19)

[0146] (20)

[0147] (21)

[0148] (22)

[0149] (23)

[0150] (24)

[0151] Formula (5) is the objective function, which minimizes the total delivery cost, which is the sum of the travel distance cost of all trucks, the use time cost of unmanned vehicles, the labor cost of drivers, and the fixed use cost of the selected distribution point. Formula (6) indicates that each customer point can only be delivered to by one vehicle once. Formula (7) indicates that when the unmanned vehicle distribution distribution point is selected, the unmanned vehicle can deliver to customers within the distribution area. Formula (8) indicates that when the unmanned vehicle distribution distribution point is selected, the truck can deliver to the distribution point. Formula (9) indicates that the unmanned vehicle cannot deliver across multiple distribution points. Formula (10) indicates that the unmanned vehicle cannot deliver to customers outside the distribution point area. Formula (11) indicates that the flow of unmanned vehicles and trucks entering and leaving the node is conserved. Formula (12) indicates that the truck departs from the distribution center and returns to the distribution center after completing the delivery. Formula (13) indicates that the unmanned vehicle departs from the distribution point and returns to the original distribution point after completing the delivery. Formula (14) indicates that each truck can be activated at most once. Formula (15) and Formula (16) respectively represent the calculation methods for the time it takes for a truck and an unmanned vehicle to arrive at the next node. Formula (17) indicates that the departure time of a vehicle at a node is equal to the sum of the arrival time at the node and the service time of the node. Formula (18) represents the change in the vehicle's load before and after serving a node. Formula (19) represents the calculation method for the remaining drivable distance when the unmanned vehicle arrives at the customer point. Formula (20) indicates that the remaining drivable distance after the unmanned vehicle delivers to a customer must be able to return to the original distribution point. Formula (21) represents the truck capacity constraint. Formula (22) indicates that the unloading volume of a truck at a distribution point is equal to the sum of the demand of all customers in the distribution point's service area. Formula (23) represents the unmanned vehicle capacity constraint. Formula (24) is the value constraint of the decision variable.

[0152] It should be noted that the Vehicle Routing Problem (VRP) is a typical NP-Hard problem, and it is difficult to find the global optimal solution using traditional exact algorithms. The location selection and path planning problem of the joint delivery of trucks and unmanned vehicles is more complex and more difficult to solve than the traditional VRP. The Adaptive Large Neighborhood Search Algorithm (ALNS) is a heuristic algorithm with flexible operation mechanism, adaptability and efficient search capabilities. It is widely used to solve various combinatorial optimization problems. Therefore, this paper designs an improved adaptive large neighborhood search algorithm (IALNS) based on the characteristics of the model to solve it. The specific ideas are as follows: (1) Considering that logistics companies face different customer needs each time they deliver, a dynamic location selection strategy for unmanned vehicle distribution points is designed. (2) In order to expand the path search range and improve the quality of the solution, a variety of destruction operators and repair operators are designed for truck paths. The destruction operator destroys the local structure of the current solution and breaks the original path restrictions, while the repair operator reconstructs the destroyed path; a variety of neighborhood search operators are designed for unmanned vehicle paths to enhance the algorithm's search capability. (3) In order to prevent the algorithm from falling into local optimality, the simulated annealing acceptance criterion is introduced to accept poor solutions with a certain probability. And by controlling the temperature parameters, the global search capability of the algorithm and the diversity of solutions are improved. The IALNS flow chart is as follows Figure 5 shown.

[0153] In specific implementation, the IALNS algorithm steps are as follows:

[0154] Step 1: Initialization.

[0155] Step 1.1: Input data such as the coordinates of the distribution center, the coordinates of the candidate distribution points for unmanned vehicle delivery, the service range of the unmanned vehicle, the customer coordinates, the truck capacity, the unmanned vehicle capacity, the customer demand, and the initial selection threshold of the distribution points for unmanned vehicle delivery.

[0156] Step 1.2: Make is the maximum number of iterations, is the number of outer layer iterations, is the number of inner layer iterations, is the optimal total delivery cost, For the optimal path, is the initial temperature, is the temperature cooling coefficient, is the length of the Markov chain. Set 、 、 、 .

[0157] Step 2: Select the location of the unmanned vehicle distribution point.

[0158] Step 2.1: Select the threshold value For each unmanned vehicle distribution point Make a selection to determine whether the threshold condition is met.

[0159] Step 2.2: Encode all the distribution points of the unmanned vehicle to form a selection state coding sequence, and record the distribution point as 1 if it is selected and 0 if it is not selected.

[0160] Step 2.3: Traverse the set of hub-distributor selection states to determine whether the selection state encoding sequence already exists. If so, find a corresponding set with the same hub-distributor selection state. Use the inner optimized path results, operator weights, and annealing temperature variables as initial variables and proceed to Step 4.2. Otherwise, proceed to Step 3.

[0161] Step 3: Construct the initial path using the greedy strategy.

[0162] Step 3.1: For all selected distribution points , filter out customer points that meet the conditions ,like , then insert the customer point into the unmanned vehicle delivery set ,in Distribution points for unmanned vehicles Customers within the service scope.

[0163] Step 3.2: Remove the unmanned vehicle delivery customer points from the set of all delivery customer points, and insert the selected distribution points to construct the truck delivery set .

[0164] Step 3.3 Calculate the unmanned vehicle delivery set The objective function value of each customer in the set (the distance from the customer to the distribution point) is calculated, and the customers in the set are sorted in ascending order according to the objective function value.

[0165] Step 3.4: Select customers from the sorted AV delivery set, from smallest to largest, and insert them into the delivery path. If the current delivery set is empty, proceed to Step 3.7; otherwise, proceed to Step 3.5.

[0166] Step 3.5 When the customer point is inserted into the current delivery path, determine whether the capacity constraint is met. If so, go to Step 3.6. Otherwise, save the original path and add a new unmanned vehicle delivery path.

[0167] Step 3.6: Determine whether the remaining driving distance of the unmanned vehicle meets the power constraint when delivering to the customer. If so, save the customer point as the current unmanned vehicle delivery route and go to Step 3.4. Otherwise, save the original route and add a new unmanned vehicle delivery route.

[0168] Step 3.7 Calculate the truck delivery set Calculate the distance between each customer and the distribution center and arrange the calculation results in ascending order.

[0169] Step 3.8 selects customers from the set in ascending order and determines whether the capacity constraint is met. If so, go to Step 3.9. Otherwise, save the original route and add a new truck delivery route.

[0170] Step 3.9 determines whether the current set is an empty set. If so, go to Step 4; otherwise, go to Step 3.8.

[0171] Step 4: Adaptive large neighborhood search.

[0172] Step 4.1: Initialize the path and let the initial solution be , the current solution is , , initialize the weight coefficient of each operator to 1, and the score and usage times of each operator to 0.

[0173] Step 4.2: Based on the operator weight, choose one of the three destruction operators (random destruction, worst cost destruction, and correlated destruction) according to the roulette wheel method to destroy the truck path and obtain the solution. .

[0174] (1) Random destruction operator: Determine the number of customers to be destroyed based on the total number of delivery customers , randomly destroying from the delivery mission until destruction indivual.

[0175] (2) Worst cost destruction: Calculate the difference between the cost of each customer point in the path after it is destroyed and the cost of the original path before it is destroyed, and destroy the results in descending order until the specified destruction amount is reached.

[0176] (3) Correlation destruction: Calculate the similarity between two customer points according to formula (25), and remove them from the highest to the lowest similarity until the specified destruction amount is reached.

[0177] (25)

[0178] in, and is the weight coefficient, Points for customers To the customer point distance, and Customer points and customer points If the customer clicks and customer points On the same path 0 if yes, 1 otherwise.

[0179] Step 4.3: According to the operator weight, select one of the three repair operators (random repair, greedy repair, and regret repair) according to the roulette wheel method to repair the truck path and obtain the solution. .

[0180] (1) Random repair: A customer point is randomly selected from the set of damaged customer points and inserted into a random position on the truck delivery path until there are no remaining customer points to be inserted.

[0181] (2) Greedy repair: Select a customer point from the set of destroyed customer points, calculate the incremental delivery cost when the customer point is inserted into different locations on different delivery routes of the truck, and select the insertion position with the smallest incremental delivery cost and insert it into that position until there are no remaining customer points to be inserted.

[0182] (3) Regret repair: Calculate the delivery cost of each customer in the damaged customer point set at the optimal insertion position and the delivery cost of the suboptimal insertion position. The difference between the costs of the two positions is used as the regret value. Select the customer point with the largest regret value and insert it into the optimal position.

[0183] Step 4.4: Use the roulette wheel method to select one of the four neighborhood search operators (Swap operator, Insert operator, 2-opt operator, or-opt operator) to obtain the delivery unmanned vehicle path. Among them, the neighborhood operator diagram is as follows Figure 6 shown.

[0184] (1) Swap operator: swaps the positions of two random customer points in the delivery path of the unmanned vehicle (e.g. Figure 6 (a));

[0185] (2) Insert operator: Randomly select two customer points on the unmanned vehicle delivery path , the customer points Insert Customer Point Afterwards (e.g. Figure 6 (b));

[0186] (3) 2-opt operator means randomly selecting 2 customer nodes , the client node With the client node The customer positions between Figure 6 (c));

[0187] (4) or-opt operator: randomly select two customer points on the unmanned vehicle delivery path , the customer points and customer points Flip and insert into the customer point Afterwards (e.g. Figure 6 (d));

[0188] Step 4.5: Calculate the sum of the delivery costs of the unmanned vehicle and the truck. Update the paths of the unmanned vehicle and the truck according to the calculation results. If , then let On this basis, if , then let 、 Otherwise, follow The acceptance criteria are updated. The acceptance criteria are as follows: If the new solution is inferior to the current solution, it is accepted according to the probability of formula (26):

[0189] (26)

[0190] in, For the new solution cost, is the current solution cost, Annealing temperature for the current iteration.

[0191] Step 4.6 updates the weights of each operator. The scores of the destruction operator and the repair operator are determined according to the solution. Specifically: (1) When the operator scores ; (2) When (1) is satisfied and When the operator scores ; (3) When and satisfy When accepting the criteria, the operator score is added ; (4) When And not satisfied When accepting the criteria, the operator score is added ; After each iteration, the operator weight is updated according to formula (27):

[0192] (27)

[0193] in, and Destruction and repair operators respectively The weights at the current and previous iterations, is the operator attenuation coefficient, and are the operator score and usage times respectively.

[0194] Step 4.7: Update annealing temperature , update the number of iterations ,like , then go to Step 4.2, otherwise, output the optimal delivery cost , optimal path , current delivery costs , Current delivery route , annealing temperature and operator weights .

[0195] Step 5: Calculate the cost of the unmanned vehicle distribution point and update the selection threshold.

[0196] Step 5.1: Determine whether the current distribution point selection code sequence already exists in the distribution point selection state set. If not, create a new distribution point selection state variable and set the optimized distribution path , annealing temperature and operator weights Otherwise, the original delivery path, annealing temperature and operator weight are updated.

[0197] Step 5.2 calculates the optimized delivery path according to the method in step 1 of the unmanned vehicle distribution point location strategy.

[0198] Step 5.3: Traverse each selected unmanned vehicle distribution point according to the formula

[0199]

[0200] Update its selection threshold.

[0201] Step 6: IALNS end condition.

[0202] ,like , then go to Step 2, otherwise, the algorithm ends and outputs the result and .

[0203] In practice, this embodiment proposes a delivery model that combines trucks and unmanned vehicles, optimizing the delivery network through the complementary functions of these two transport vehicles. For high-density delivery scenarios, such as urban communities, unmanned vehicles leverage their advantages in load capacity, endurance, and continuous operation for end-to-end delivery, helping to reduce manual labor and improve delivery efficiency.

[0204] Compared with the fixed hierarchical method of traditional hierarchical distribution, this embodiment proposes that trucks can deliver to discrete customer points along the way while heading to a distribution point. This method is more flexible in the selection of delivery points. A small increase in truck routes can replace the longer routes of unmanned vehicle delivery, effectively reducing driving distance and expanding the space for route optimization.

[0205] This embodiment proposes a dynamic location selection strategy for autonomous vehicle hubs based on delivery demand. Based on different customer distribution characteristics and analyzing the structural changes in delivery demand in the terminal area, it adaptively selects hub activation status. This effectively addresses the resource mismatch problem inherent in traditional fixed hub layouts, achieving an optimal match between facility resources and delivery demand.

[0206] This embodiment designs an improved adaptive large neighborhood search algorithm, employs multiple neighborhood operators to improve search efficiency, and introduces a simulated annealing acceptance criterion to enhance global search capabilities. This effectively optimizes delivery routes for trucks and unmanned vehicles, minimizing delivery costs.

[0207] In the specific implementation, the corresponding algorithm in the embodiment is also subjected to a case experiment.

[0208] 1. Experimental Setup

[0209] Since there is currently no standard test case library for location selection and route planning of joint delivery by trucks and unmanned vehicles, this experiment generates test cases based on the Solomon case. The generated cases include three distribution types: cluster distribution (C1, C2), random distribution (R1, R2) and mixed distribution (RC1, RC2). The service time for customers and distribution points in each case is set to 2 minutes. In addition, to meet the test requirements of this article, the following data is supplemented: Each case is based on the coordinates 、 、 and As candidate distribution points for unmanned vehicle delivery, the service area radius of each candidate distribution point is 12km.

[0210] The model parameters are set as follows: 、 、 、 、 Yuan / km, Yuan / min, Yuan / min, The IALNS parameters are set as follows: , , , , , , , , , .

[0211] IALNS was programmed with Matlab 2023a and ran on a microcomputer with a 12th Gen Intel(R) Core(TM) i7-12700H@2.30 GHz CPU and 16 GB of RAM.

[0212] 2 Experimental analysis

[0213] 2.1 Multi-type case experiments

[0214] Different types of examples are used for experiments, and the calculation results are shown in Table 1, where: IN represents the example name, TC represents the total delivery cost (unit: yuan), TT represents the total driving time of trucks and unmanned vehicles (unit: min), TRC represents the driving cost of trucks (unit: yuan), UC represents the use cost of unmanned vehicles (unit: yuan), LC represents the labor cost (unit: yuan), DPC represents the fixed use cost of the unmanned vehicle distribution point (unit: yuan), UD represents the driving distance of the unmanned vehicle (unit: km), DPN represents the number of selected distribution points (unit: piece), and RT represents the algorithm running time (unit: second).

[0215] According to the results in Table 1, (1) From the values ​​of TC, TRC, UC, LC and DPC, it can be seen that the driving cost of trucks, the cost of using unmanned vehicles, the labor cost and the fixed cost of using unmanned vehicle distribution points account for 68.28%, 4.93%, 21.41% and 5.38% of the total distribution cost respectively. This shows that the driving cost of trucks and labor cost are the main factors affecting logistics costs. Therefore, logistics companies should rationally plan the driving routes of trucks, optimize the distribution mode, minimize the driving cost of trucks, reduce the number of customer points directly delivered by trucks, and control the labor cost of truck distribution. (2) From the values ​​of TC, UC and UD, it can be seen that the average driving mileage of unmanned vehicles reaches 245.33km. Compared with the higher driving cost and labor cost of trucks, unmanned vehicles only consume electricity during the distribution process and do not require manpower. The distribution cost is significantly reduced, accounting for only 4.93% of the total distribution cost. This shows that the use of unmanned vehicle distribution is both economical and environmentally friendly. Therefore, logistics companies should give priority to unmanned vehicle distribution, which not only helps to reduce logistics costs, but also improves logistics distribution performance and promotes the sustainable development of the logistics industry. (3) From the values ​​of UC, DPC, and DPN, it can be seen that the cost of using unmanned vehicles and the fixed cost of using distribution points increase with the increase in the number of distribution points selected, indicating that the number of distribution points selected will affect the total distribution cost within the distribution point area when using unmanned vehicles for delivery. Therefore, logistics companies should reasonably select unmanned vehicle distribution distribution points based on the actual distribution of delivery customers. (4) From the value of DPN, it can be seen that under different customer distribution types, the number of selected unmanned vehicle distribution distribution points is different. In the clustered distribution type C example, the number of selected unmanned vehicle distribution distribution points is the largest; in the randomly distributed type R example, the number of selected unmanned vehicle distribution distribution points is the least; in the mixed distribution type RC example, the number of selected unmanned vehicle distribution distribution points is between the R and C examples. This shows that the dynamic location selection strategy for unmanned vehicle distribution distribution points designed in this paper can flexibly select locations according to the distribution of different types of customers, and is reasonable and feasible. (5) From the value of RT, we can see that the average running time of the algorithm is 370.63, which shows that IALNS can solve various types of large-scale examples within a reasonable time and has high practicality and operating efficiency.

[0216] Table 1 Experimental results of different examples

[0217]

[0218] The location selection and path planning diagrams of the unmanned vehicle for cases C101, C201, R101, and RC202 are shown in the figure below. Figure 7 shown.

[0219] Depend on Figure 7 It can be seen that:

[0220] Unmanned vehicles are responsible for delivering to customers within the selected distribution point area, and trucks are responsible for delivering to customers within the selected distribution point area and the unselected distribution point area.

[0221] Figure 7 (a) and Figure 7 In (b), the customers are clustered, and the number of selected distribution points is four. Three trucks are used in each distribution center. In the C1 case, one unmanned vehicle is used at each distribution point. In the C2 case, only one unmanned vehicle is used at one distribution point, and the other three distribution points each use two unmanned vehicles. Figure 7 In (c), customers are randomly distributed, two distribution points are selected, two trucks are used in the distribution center, and one unmanned vehicle is used in each selected distribution point. Figure 7 In (d), customers are mixed, and three distribution points are selected. The distribution center uses three trucks, and each selected distribution point uses one unmanned vehicle. This demonstrates that the dynamic location selection strategy for unmanned vehicle distribution distribution points can dynamically select unmanned vehicle distribution distribution points based on changes in customer demand by comparing the costs of truck delivery and unmanned vehicle delivery within the distribution point area. When a distribution point has high customer demand and low unmanned vehicle delivery costs within the distribution point area, that distribution point is prioritized for deployment; when a distribution point has low customer demand and low truck delivery costs within the distribution point area, that distribution point is not selected. Figure 7 The site selection and path planning results of different types of examples show that the algorithm in this paper can dynamically select the location of distribution points according to customer distribution and reasonably plan the paths of trucks and unmanned vehicles, which is reasonable and feasible.

[0222] 2.2 Comparative experiment between dynamic location selection strategy and fixed location selection strategy

[0223] Using multiple case studies, we conducted comparative experiments comparing dynamic location selection strategies with fixed location selection strategies (a fixed location selection strategy involves activating all pre-defined autonomous vehicle delivery hubs, regardless of customer demand). The experimental results are shown in Table 2. UCT represents the sum of the autonomous vehicle usage cost and the fixed usage cost of the hubs (collectively referred to as the autonomous vehicle delivery cost), UVN represents the number of autonomous vehicles in use, and NU represents the total number of customers within the selected autonomous vehicle delivery hubs. All other symbols have the same meanings as in Table 1.

[0224] From the results in Table 2, we can see that: (1) According to the results of TC, UCT and DPN, the total distribution cost of the dynamic location strategy in the R type and RC type cases is significantly lower than that of the fixed location strategy, with an average saving of 3.76%, and the average saving of the unmanned vehicle distribution cost is 21.84%. This shows that the dynamic location strategy can scientifically enable distribution points according to customer needs, avoid selecting unnecessary distribution points, and effectively reduce distribution costs. In contrast, the fixed location strategy selects all distribution points. Even if the demand in some areas is low, unmanned vehicle distribution will still be used, resulting in an increase in the number of distribution areas and high operating costs. (2) According to DPN and NU, it can be seen that in the C type case, the customer distribution in each distribution point area is relatively concentrated, and the number of distribution points selected by the dynamic location strategy and the fixed location strategy is the same. This shows that in the case where the customer distribution in the distribution point area is concentrated, the cost of using unmanned vehicle delivery is lower. In the R and RC type cases, the number of customers delivered by unmanned vehicles is only slightly reduced when the dynamic location selection strategy is adopted compared with the fixed location selection strategy. At this time, the number of selected distribution points is also reduced accordingly. This shows that the number of customers in the unselected distribution point area is small. The operating cost of continuing to use unmanned vehicle delivery is higher than that of truck delivery. Therefore, only some distribution points with high customer demand are selected. (3) In general, the dynamic location selection strategy can respond more accurately to changes in customer demand, reasonably activate unmanned vehicle distribution distribution points, effectively reduce unnecessary distribution point operating costs, and improve resource utilization efficiency. Therefore, when logistics companies select distribution points, they should adopt a dynamic location selection strategy to reasonably activate distribution points for distribution according to the actual situation of customer demand in the distribution point area, ensure that resource allocation matches demand, effectively control logistics costs, and improve logistics distribution performance.

[0225] Table 2 Experimental results of dynamic location selection strategy and fixed location selection

[0226]

[0227] 2.3 Comparative Experiment between Unmanned Vehicle Delivery and Traditional Vehicle Delivery

[0228] We conducted comparative experiments on a multi-case model using a truck-and-unmanned vehicle delivery model and a traditional vehicle delivery model (where all customer deliveries are made by truck). The experimental results are shown in Table 3, where TR represents truck travel time, TS represents human service time, VN represents the number of trucks used, CSR represents the cost savings ratio for total delivery costs under different delivery models, and TSR represents the time savings ratio for the sum of truck travel time and human service time under different delivery models. All other symbols have the same meanings as in Table 1.

[0229] Table 3 Experimental results of truck and unmanned vehicle delivery modes and traditional vehicle delivery modes

[0230]

[0231] From the experimental results in Table 3, we can see that: (1) According to the results of TC and CSP, the total delivery cost of the combined delivery mode of trucks and unmanned vehicles is significantly lower than that of the traditional vehicle delivery strategy. Among them, the total delivery cost is saved by 8.38% at the highest, 4.51% at the lowest, and 7.49% on average. Because unmanned vehicle delivery does not require labor costs, the delivery cost is relatively low. In the terminal delivery scenario of short-distance delivery and concentrated customer demand distribution, unmanned vehicles can effectively replace traditional vehicles to complete delivery tasks and reduce overall delivery costs. In contrast, traditional vehicle delivery requires manual driving and manual service, and its operating costs include not only vehicle driving costs but also labor costs. It shows obvious disadvantages in terminal delivery that requires frequent customer service. Therefore, logistics companies should combine the specific characteristics of the delivery scenario and reasonably adopt unmanned vehicle delivery on the basis of truck delivery, make full use of the cost advantages of unmanned delivery, and reduce the total delivery cost. (2) From the results of TR, TS and TSR, when the truck and unmanned vehicle delivery mode is adopted, the truck driving time and manual service time are significantly shorter than those of the traditional vehicle delivery mode. Specifically, the combined time savings for these two items reached a maximum of 24.69% in the C1 case, and a minimum of 15.53% in the R2 case, with an average saving of 19.03%. This demonstrates that the combined delivery model of trucks and autonomous vehicles can effectively reduce redundancy in the logistics and distribution process, reducing truck usage time by flexibly assigning autonomous vehicle tasks. Given the high cost of human resources, logistics companies can significantly reduce their reliance on manual labor, alleviate operational pressure, and further improve logistics and distribution performance by adopting a combined delivery strategy of trucks and autonomous vehicles.

[0232] 2.4 Algorithm Comparison Experiment

[0233] To verify the effectiveness of the IALNS algorithm, a simulated annealing algorithm (SAA) was designed to solve the location and routing planning of joint delivery between trucks and unmanned vehicles. The SAA was set to an initial temperature of 100, a cooling coefficient of 0.92, a Markov chain length of 300, and 800 iterations. Comparative experiments were conducted using multiple examples. The experimental results are shown in Table 4. Where TD represents the total distance traveled by the vehicle, DSR represents the total distance saved, and the remaining symbols have the same meanings as in Table 1.

[0234] Table 4 Comparison results of different optimization algorithms

[0235]

[0236] From the experimental results in Table 4, we can see that: (1) According to the values ​​of TC and TD, the total delivery cost and total driving distance solved by IALNS under different types of cases are better than SAA. Among them, the total delivery cost is saved by 9.63% at the highest, 4.99% at the lowest, and 6.84% on average, while the total driving distance is saved by 5.28% at the highest, 1.88% at the lowest, and 3.05% on average. This shows that IALNS can effectively reduce delivery costs, shorten delivery distances, and reduce delivery time. (2) According to the value of RT, IALNS significantly shortens the solution time when solving C-type and RC-type cases compared with SAA. Although IALNS takes longer than SAA to solve R-type cases, it is still within a reasonable time range and can save an average of 7.08% of time overall. Therefore, IALNS is reasonable, reliable, and effective in solving the location selection and path planning of joint delivery of trucks and unmanned vehicles.

[0237] 2.5 Sensitivity analysis

[0238] (1) Sensitivity analysis of the service scope of the distribution point

[0239] To verify the impact of the hub service range on the routing of trucks and autonomous vehicles, the hub service radius was varied within the range [10, 18] with a step size of 2, keeping other parameters constant. Experiments were conducted using cases C206, R106, and RC106. The experimental results are shown in Table 5, where R represents the autonomous vehicle service radius at the hub, UT represents the autonomous vehicle travel time, and the meanings of the remaining symbols are the same as in Tables 1 and 2.

[0240] The results in Table 5 show that: (1) According to the values ​​of TC, UT and UCT, under different types of cases, as the service range of the distribution point gradually expands, the total distribution cost shows a significant downward trend, while the delivery time and delivery cost of the unmanned vehicle show an upward trend. This is because the expansion of the service range of the distribution point increases the number of customers delivered by the unmanned vehicle, and more delivery tasks are undertaken by the unmanned vehicle, while the delivery tasks of the truck are reduced accordingly. Since the unmanned vehicle has lower driving cost and service cost, the reduction in the delivery cost of the truck is significantly greater than the increase in the delivery cost of the unmanned vehicle, so the total distribution cost gradually decreases overall. In addition, as the service range of the unmanned vehicle expands, the unmanned vehicle needs to serve more customers, which inevitably leads to an increase in the driving time and delivery cost of the unmanned vehicle. (2) Different types of cases have different sensitivities to changes in the service range of the distribution point. For the C2 and R1 cases: when R (the service radius of the unmanned vehicle at the distribution point) increases from 10 to 16, the total distribution cost decreases significantly, while when R increases from 16 to 18, the decrease in the total distribution cost slows down significantly. For the RC1 example, when R increases from 14 to 16, the decrease in total delivery costs is relatively small, while in other ranges, the decrease in total delivery costs is more significant. Therefore, when expanding the service scope of distribution points, logistics companies should comprehensively consider customer needs and actual operational conditions, evaluate the necessity of expansion, and avoid excessive expansion of service scope that incurs additional delivery costs and time.

[0241] Table 5 Experimental results of sensitivity analysis of the size of the distribution area

[0242]

[0243] (2) Analysis of the sensitivity of the time cost of unmanned vehicles

[0244] In order to study the impact of the time cost of unmanned vehicles on the location planning of unmanned vehicle distribution points, the time cost of unmanned vehicles is adjusted in steps of 0.1 between [0.1, 0.5], while keeping other parameters unchanged.

[0245] The experiment was conducted using case RC107. The experimental results are shown in Table 6, where P represents the time cost of the unmanned vehicle. The meanings of the other symbols are the same as in Table 1.

[0246] Table 6 Experimental results of sensitivity analysis of unmanned vehicle usage costs

[0247]

[0248] The results in Table 6 show that: (1) According to the values ​​of P, UC, DPC, and DPN, as the time cost of unmanned vehicles increases, the number of selected distribution points and the fixed cost of unmanned vehicle use show a trend of gradual reduction. Because when the time cost of unmanned vehicles increases, the distribution point areas that originally had distribution cost advantages gradually lose this advantage, resulting in the cost of unmanned vehicle distribution at some distribution points being higher than that of truck distribution, and thus no longer being selected. This shows that under the same customer demand distribution, the selection of unmanned vehicle distribution points will be affected by the time cost of unmanned vehicles. (2) According to the values ​​of P and TC, when P increases from 0.1 to 0.4, the total distribution cost increases more significantly; when P increases from 0.4 to 0.5, the total distribution cost increases more slowly. Therefore, in actual distribution planning, the time cost of unmanned vehicles should be fully considered, the number of unmanned vehicle distribution points activated and the proportion of unmanned vehicles deployed should be reasonably selected, the distribution plan should be optimized, the total distribution cost should be reduced, and the distribution performance should be improved.

[0249] (3) Sensitivity analysis of fixed usage costs of distribution points

[0250] To study the impact of fixed usage costs on the location planning of autonomous vehicle hubs, we kept other parameters constant and varied the fixed usage costs of individual hubs within the range [30, 90] with a step size of 15. Experiments were conducted using case RC108. The experimental results are shown in Table 7. Z represents the fixed usage cost of a single hub, and the meanings of the other symbols are the same as in Table 1.

[0251] The results in Table 7 show that: (1) According to the values ​​of Z and DPN, as the fixed cost of a single distribution point increases, the number of distribution points selected decreases. Because the increase in the fixed cost of a single distribution point leads to an increase in the total distribution cost of the distribution point area, if the number of customers in the distribution point area is small, it is difficult to expand the cost advantage through unmanned vehicle delivery, making the unmanned vehicle delivery cost of the entire distribution point area high, and ultimately causing the originally selected distribution point to no longer be selected. (2) According to the values ​​of Z, TC, UC and DPN, when Z increases from 45 to 60, although the cost of unmanned vehicle use decreases due to the reduction in the number of selected distribution points, the total distribution cost increases. This is because the distribution point area originally delivered by unmanned vehicles is converted to truck delivery, and although the truck delivery cost is lower than the unmanned vehicle delivery cost in the area, it is still higher than the unmanned vehicle use cost, so the total cost increases. In summary, the fixed cost of distribution points will affect the selection of distribution points. Therefore, logistics companies should pay attention to the fixed usage costs of distribution points when selecting locations. If the fixed usage costs of distribution points are too high and the number of customers in the distribution point area is small, it is not advisable to select this distribution point.

[0252] Table 7 Experimental results of sensitivity analysis of fixed usage cost of distribution points

[0253]

[0254] This embodiment, upon receiving a delivery task, obtains candidate collection and distribution points for unmanned vehicles and their service areas based on the delivery task. Based on customer demand for the delivery task, it dynamically adjusts the collection and distribution point selection threshold to determine whether each candidate collection and distribution point is enabled. For enabled collection and distribution points, it optimizes the combined delivery path of trucks and unmanned vehicles using an improved adaptive large neighborhood search algorithm to minimize the total delivery cost. The algorithm then calculates and outputs the total delivery cost, the optimal path, and the collection and distribution point selection status. For the combined delivery model of trucks and unmanned vehicles, a dynamic collection and distribution point selection strategy for unmanned vehicles based on delivery demand is proposed to flexibly allocate delivery tasks, reduce manual workload, and improve delivery efficiency.

[0255] In addition, an embodiment of the present application also proposes a computer-readable storage medium, which stores a program for site selection and path decision-making for joint delivery of trucks and unmanned vehicles. When the program for site selection and path decision-making for joint delivery of trucks and unmanned vehicles is executed by a processor, the steps of the method for site selection and path decision-making for joint delivery of trucks and unmanned vehicles as described above are implemented.

[0256] Reference Figure 8 , Figure 8 This is a structural block diagram of the first embodiment of the site selection and route decision-making device for joint delivery by trucks and unmanned vehicles in this application.

[0257] like Figure 8 As shown, the location selection and route decision-making device for the joint delivery of trucks and unmanned vehicles proposed in the embodiment of the present application includes:

[0258] The data acquisition module 10 is used to obtain candidate collection and distribution points for unmanned vehicle delivery and their service ranges according to the delivery task when receiving the delivery task;

[0259] A judgment module 20 is used to obtain customer demand based on the delivery task and dynamically adjust the collection and distribution point selection threshold to determine whether each candidate collection and distribution point is enabled;

[0260] Algorithm module 30, for enabled distribution points, optimizes the joint delivery routes of trucks and unmanned vehicles through an improved adaptive large neighborhood search algorithm to minimize the total delivery cost;

[0261] The calculation module 40 is used to calculate and output the total distribution cost, the optimal path and the distribution point selection status, wherein the total distribution cost includes the truck driving cost, the unmanned vehicle usage time cost, the driver labor cost and the distribution point fixed usage cost.

[0262] It should be understood that the above is only an example and does not constitute any limitation to the technical solution of the present application. In specific applications, technicians in this field can make settings as needed, and the present application does not impose any restrictions on this.

[0263] This embodiment, upon receiving a delivery task, obtains candidate collection and distribution points for unmanned vehicles and their service areas based on the delivery task. Based on customer demand for the delivery task, it dynamically adjusts the collection and distribution point selection threshold to determine whether each candidate collection and distribution point is enabled. For enabled collection and distribution points, it optimizes the combined delivery path of trucks and unmanned vehicles using an improved adaptive large neighborhood search algorithm to minimize the total delivery cost. The algorithm then calculates and outputs the total delivery cost, the optimal path, and the collection and distribution point selection status. For the combined delivery model of trucks and unmanned vehicles, a dynamic collection and distribution point selection strategy for unmanned vehicles based on delivery demand is proposed to flexibly allocate delivery tasks, reduce manual workload, and improve delivery efficiency.

[0264] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this application. In actual applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of this embodiment scheme, and no restrictions are imposed here.

[0265] In addition, for technical details not fully described in this embodiment, please refer to the site selection and route decision method for joint delivery of trucks and unmanned vehicles provided in any embodiment of this application, and will not be repeated here.

[0266] In addition, it should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0267] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0268] Through the above description of the embodiments, those skilled in the art will clearly understand that the above-mentioned embodiments and methods can be implemented using software plus the necessary general-purpose hardware platform. Of course, hardware can also be used, but in many cases the former is a more preferred embodiment. Based on this understanding, the technical solution of this application, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, a magnetic disk, or an optical disk) and includes several instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of this application. The above are only preferred embodiments of this application and do not limit the scope of the patent application. Any equivalent structure or equivalent process transformation made using the contents of this application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the scope of patent protection of this application.

Claims

1. A location selection and routing decision method for joint delivery of trucks and unmanned vehicles, characterized by: include: Upon receiving a delivery task, obtaining candidate collection and distribution points for unmanned vehicle delivery and their service ranges according to the delivery task; Obtain customer demand based on the delivery task, and dynamically adjust the distribution point selection threshold to determine whether each candidate distribution point is enabled; For enabled distribution points, an improved adaptive large neighborhood search algorithm is used to optimize the joint delivery routes of trucks and unmanned vehicles to minimize the total delivery cost. Calculate and output the total delivery cost, optimal route, and distribution point selection status, where the total delivery cost includes truck driving costs, unmanned vehicle usage time costs, driver labor costs, and fixed distribution point usage costs; The improved adaptive large neighborhood search algorithm includes: The destruction operators are used to perform local destruction on the truck routes, including random destruction, worst cost destruction and correlated destruction. Use repair operators to reconstruct the damaged path, including random repair, greedy repair and regret repair; Design neighborhood search operators for autonomous vehicle paths, including Swap, Insert, 2-opt, and or-opt operators; The simulated annealing acceptance criterion is introduced to accept poor solutions in a probabilistic manner to prevent the algorithm from falling into local optimality. The step of reconstructing the damaged path using a repair operator includes: Obtain the operator weight and select one of the three repair operators according to the roulette wheel method to repair the truck path and obtain the solution. ; Among them, random repair includes: randomly selecting a customer point from the set of damaged customer points and inserting it into a random position of the truck delivery path until there are no remaining customer points to be inserted; Greedy repair includes: selecting a customer point from the set of damaged customer points, calculating the incremental delivery cost when the customer point is inserted into different locations on different delivery routes of the truck, and selecting the insertion location with the smallest incremental delivery cost and inserting it there until there are no remaining customer points to be inserted; Regret repair includes: calculating the delivery cost of each customer in the set of damaged customer points at the optimal insertion position and the delivery cost of the suboptimal insertion position, taking the difference between the two costs as the regret value, and selecting the customer point with the largest regret value and inserting it into the optimal position; The step of designing a neighborhood search operator for the unmanned vehicle path includes: According to the roulette method, one of the four neighborhood search operators is selected to obtain the delivery unmanned vehicle path. ; Among them, the Swap operator exchanges the positions of two random customer points in the unmanned vehicle delivery path; The Insert operator randomly selects two customer points on the delivery path of the unmanned vehicle. , the customer points Insert Customer Point after; The 2-opt operator randomly selects two client nodes , the client node With the client node The customer positions between them are in reverse order; The or-opt operator randomly selects two customer points on the unmanned vehicle delivery path. , the customer points and customer points Flip and insert into the customer point after.

2. The method according to claim 1, characterized in that The dynamic adjustment of the collection and distribution point selection threshold specifically includes: Calculate the difference between the delivery cost of unmanned vehicles when the distribution point is enabled and the delivery cost of trucks when it is not enabled; Dynamically update the selection threshold based on the difference. If the delivery cost of the unmanned vehicle is lower, then increase the selection threshold to increase the probability of the distribution point being activated. Otherwise, decrease the selection threshold. The normal distribution function is used to determine whether the collection and distribution point is enabled, and a collection and distribution point selection state coding sequence is generated.

3. The method according to claim 1, characterized in that The similarity of the destruction operator is calculated as: in, and is the weight coefficient, Points for customers To the customer point distance, and Customer points and customer points If the customer orders and customer points On the same path If it is 0, otherwise it is 1, and the similarities are removed from large to small until the specified number of destructions is reached.

4. The method according to claim 1, wherein The mathematical model of the total distribution cost is: in, is the distance traveled by the truck, The time cost of using the autonomous vehicle, Labor costs for drivers, The usage cost of the distribution point is fixed, and the path flow conservation and capacity constraints are met.

5. A location selection and routing decision-making device for joint delivery of trucks and unmanned vehicles, characterized by: Executing the method according to claim 1, comprising: A data acquisition module is used to obtain candidate collection and distribution points for unmanned vehicle delivery and their service ranges based on the delivery task when a delivery task is received; A judgment module is used to obtain customer demand based on the delivery task and dynamically adjust the collection and distribution point selection threshold to determine whether each candidate collection and distribution point is enabled; The algorithm module optimizes the joint delivery routes of trucks and unmanned vehicles for enabled distribution points through an improved adaptive large neighborhood search algorithm to minimize the total delivery cost. The calculation module is used to calculate and output the total distribution cost, the optimal route and the distribution point selection status, wherein the total distribution cost includes the truck driving cost, the unmanned vehicle usage time cost, the driver labor cost and the fixed usage cost of the distribution point.

6. A computer device, characterized in that: The device comprises: a memory and a processor, wherein the processor executes the method according to any one of claims 1 to 4 when running computer instructions stored in the memory.

7. A computer-readable storage medium, characterized in that The method comprises instructions which, when executed on a computer, cause the computer to execute the method according to any one of claims 1 to 4.

Images