A method and system for optimizing the delivery path of a drone and a truck

Abstract

The application discloses a kind of unmanned aerial vehicle and truck logistics distribution path optimization method and system, truck carries unmanned aerial vehicle and express delivery departs from distribution center, flies unmanned aerial vehicle or distributes express delivery to fixed point, unmanned aerial vehicle can complete multiple customer points distribution task each time take off, truck does not need to wait unmanned aerial vehicle to return to base, but according to path planning scheme directly goes to next cluster cluster center point and unmanned aerial vehicle meet, mutual coordination and cooperation of truck and unmanned aerial vehicle in distribution process, improve logistics distribution efficiency, and then reduce final distribution cost.

Description

Technical Field

[0001] This invention relates to the field of logistics and distribution technology, specifically to a method and system for optimizing logistics and distribution routes using drones and trucks. Background Technology

[0002] Due to the low population density, low logistics demand per unit area, complex delivery routes, and long delivery times in rural areas, the "last mile" delivery is particularly difficult. Residents usually have to go to a courier point far from their home to pick up their packages, and cannot enjoy the efficiency and convenience of "last mile" delivery. Therefore, it is necessary to improve and innovate the traditional delivery model.

[0003] Compared to the traditional human-vehicle delivery model, where delivery personnel drive logistics vehicles to delivery addresses, drone delivery is more efficient and flexible. It avoids the challenges of traffic congestion and complex road layouts common in traditional "last-mile" delivery. Furthermore, the dispersed distribution of households and the lower density of buildings and populations in rural areas provide favorable airspace conditions for low-altitude drone flights. However, considering the limited flight radius and loading constraints of drones, a truck-based approach is adopted. Drones are transported to designated locations for launch and retrieval, enabling drone-truck joint delivery. This effectively compensates for the limitations imposed by drones on individual delivery methods, such as battery and loading constraints. More importantly, drone delivery reduces human-to-human contact.

[0004] As early as 2013, e-commerce giant Amazon announced its idea of ​​using drones to deliver orders. Subsequently, logistics giants such as DHL, Federal Express, and UPS launched drone delivery experiments, and domestic companies such as SF Express and JD.com also actively explored the field, gaining widespread recognition in the industry. However, existing drone delivery is usually only used for single-point delivery, and its advantages are ultimately limited by its small payload capacity and short battery life. The academic community has long recognized the need to combine drones with larger vehicles (trucks) to improve the efficiency and effectiveness of last-mile delivery.

[0005] Unlike traditional vehicle routing problems, optimizing the routes for joint drone and truck delivery requires considering the characteristics of drone operation, such as takeoff and landing points, battery replacement, and restocking. These factors make the solution more complex. Furthermore, existing research on vehicle-mounted drone delivery typically involves a single drone taking off to serve only one customer location, and the truck must wait for the drone to return. This undoubtedly reduces the utilization rate of drone capacity and overall delivery efficiency. Therefore, a better optimization strategy is urgently needed for current drone-truck joint delivery routes. Summary of the Invention

[0006] In view of the above-mentioned shortcomings in the prior art, the present invention provides a method and system for optimizing the logistics delivery route of drones and trucks, which solves the problem of low efficiency in the joint delivery of drones and trucks.

[0007] To achieve the aforementioned objectives, the present invention employs the following technical solution: a method for optimizing delivery routes using drones and trucks, comprising the following steps:

[0008] S1. Based on the location of the delivery customer points, perform clustering processing on the delivery customer points to obtain different clusters;

[0009] S2. Using the central node of the cluster as the access node for trucks and the customer points within the cluster as the access nodes for drones, the routes of drones and trucks are planned to obtain the optimal routes for drones and trucks in the logistics delivery process.

[0010] Further, step S1 includes the following sub-steps:

[0011] S11. Within the range of the delivery customer's location, randomly generate multiple cluster centers;

[0012] S12. Calculate the distance from each customer point to all cluster centers;

[0013] S13. Determine whether the minimum distance from each customer point to all cluster centers satisfies the maximum radius constraint of the nearest cluster. If yes, the customer point is added to the nearest cluster to obtain a temporary cluster, and the process jumps to step S14. If no, the process jumps to step S15.

[0014] S14. Determine whether the total weight of all customer points in the temporary cluster meets the maximum load constraint of the drone. If yes, the temporary cluster is a cluster. If not, proceed to step S15.

[0015] S15. Generate a new cluster center and jump to step S12. Continue until all customer points have corresponding clusters. Then the clustering process of the delivery customer points is completed, and multiple clusters are obtained.

[0016] The beneficial effects of the above-mentioned further solution are as follows: This invention improves the existing K-means algorithm by implementing a self-optimizing process for determining the K value. It restricts customer points from entering the cluster by using two constraints: the maximum radius of the cluster and the maximum payload of the UAV. It covers all customer points by continuously increasing the number of cluster centers.

[0017] Furthermore, step S2 includes the following sub-steps:

[0018] S21. Use the distribution center as the starting access node for trucks and the central node of the cluster as the delivery access node for trucks.

[0019] S22. Use all customer points within a cluster as access nodes for a single drone takeoff.

[0020] S23. Use a genetic algorithm to process the visited nodes to obtain the optimal path for the drone and truck in the logistics delivery process.

[0021] Furthermore, step S23 includes the following sub-steps:

[0022] S2301. Use natural numbers to encode drones, trucks, and access nodes;

[0023] S2302. Based on the encoded numbers, randomly generate an initial population of size Nind;

[0024] S2303. Inspect and remove unfeasible initial individuals;

[0025] S2304. Calculate the fitness of all individuals retained from the initial population;

[0026] S2305. Store high-fitness individuals in the cache unit;

[0027] S2306. Select Nsel chromosomes from the parent generation using the roulette wheel method to obtain the initial offspring individuals;

[0028] S2307. Based on the crossover probability, select chromosomes from the initial offspring individuals for crossover operation to generate the second part of offspring individuals;

[0029] S2308. Based on the mutation probability, select chromosomes from the second batch of offspring individuals for mutation operations to generate the third batch of offspring individuals;

[0030] S2309. Sort the individuals in the initial population in step S2304 according to fitness from low to high, select individuals with fitness from low to high (Nind-Nsel) as cache individual units, and merge the cache individual units with the third part of the offspring individuals to form a complete population of size Nind, i.e., the new offspring population.

[0031] S2310. Determine whether the number of iterations has reached the maximum number of iterations. If yes, decode the new offspring population to obtain the optimal path for the drone and the truck. If no, use the new offspring population as the parent population and jump to step S2304.

[0032] Furthermore, the objective function for calculating the fitness of all individuals retained in the initial population in step S2304 is:

[0033] minZ = F1 + F2 + F3

[0034] Where Z represents the individual's fitness, F1 represents the truck delivery cost, F2 represents the drone delivery cost, and F3 represents the time cost.

[0035] Furthermore, the formula for calculating the truck delivery cost is as follows:

[0036]

[0037]

[0038] Where F1 is the truck delivery cost, c t For truck start-up costs, c T Let x be the cost per unit distance traveled by the truck, i be the i-th access node of the truck, j be the j-th access node of the truck, and x be the cost per unit distance traveled by the truck. ij Let d be the state parameter of the truck from the i-th access node to the j-th access node. ij Let N be the distance from the i-th access node to the j-th access node of the truck, and let N be the number of nodes in the road network, including the distribution center N0 and the number of customer points with delivery needs in the cluster. c Truck delivery access node N f .

[0039] Furthermore, the formula for calculating drone delivery costs is as follows:

[0040]

[0041]

[0042] Where F2 is the drone delivery cost, M is the number of clusters with more than 1 customer points, and c d For the startup cost of drones, c D Let N be the unit distance cost of drone flight, i be the i-th customer point in the cluster visited by the drone, j be the j-th customer point in the cluster visited by the drone, and N be the total distance cost. c Let y′ represent the number of customer points with delivery needs within the cluster. ijf Let d be the state parameters of the drone from the i-th customer point to the j-th customer point. ij Let be the distance between the i-th customer point and the j-th customer point for the drone.

[0043] Furthermore, the formula for calculating the time cost is as follows:

[0044] F3 = w T +w D

[0045]

[0046]

[0047] Where F3 is the time cost, w T For truck waiting time, w D Let c0 be the time penalty cost incurred due to breach of customer service hours, c0 be the unit time cost of the drone waiting for the truck, and N be the number of nodes in the road network, including the distribution center N0 and the number of customer points with delivery needs in the cluster N. c Truck delivery access node N f k T Let k be the set of nodes accessed by trucks. D Let k be the set of nodes accessed by drones, k be the number of nodes accessed by both trucks and drones, and x be the number of nodes accessed by both. ij Let d be the state parameter of the truck from the i-th access node to the j-th access node. ij Let N be the distance from the i-th access node to the j-th access node of the truck. c d represents the number of customer points with delivery needs within the cluster. ij Let y′ be the distance from the i-th access node to the j-th access node for the drone. ijf Let v be the state parameters of the drone from the i-th customer point to the j-th customer point. T v is the speed of the truck. D Let e ​​be the drone's flight speed, and e be the waiting penalty coefficient. i f is the delivery start time for the i-th customer point. i Let t be the delivery deadline for the i-th customer point. i Let f be the time it takes for the drone to reach the i-th customer point, and f be the lateness penalty coefficient.

[0048] In summary, the beneficial effects of this invention are as follows: a method and system for optimizing the logistics delivery route of drones and trucks, in which trucks carrying drones and express packages depart from the distribution center and go to a fixed point to launch the drones or deliver the packages. Each time the drone takes off, it can complete delivery tasks for multiple customer points. The truck does not need to wait for the drone to return to its original location, but instead goes directly to the next cluster center point to meet the drone according to the route planning scheme. The mutual coordination and cooperation between the truck and the drone in the delivery process improves the efficiency of logistics delivery, thereby reducing the final delivery cost. Attached Figure Description

[0049] Figure 1A flowchart of a method for optimizing delivery routes using drones and trucks;

[0050] Figure 2 This is a diagram illustrating the combined delivery of drones and trucks.

[0051] Figure 3 This is a schematic diagram illustrating the construction of the initial solution of the genetic algorithm of the present invention;

[0052] Figure 4 This is a schematic diagram of the crossover operator in the genetic algorithm of this invention;

[0053] Figure 5 This is a schematic diagram of the mutation operator in the genetic algorithm of the present invention;

[0054] Figure 6 This is a schematic diagram of the customer point clustering results of the present invention;

[0055] Figure 7 This is a schematic diagram illustrating the iterative convergence of the method of the present invention. Detailed Implementation

[0056] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0057] Example 1:

[0058] like Figure 1 As shown, a method for optimizing delivery routes using drones and trucks includes the following steps:

[0059] S1. Based on the location of the delivery customer points, perform clustering processing on the delivery customer points to obtain different clusters;

[0060] Step S1 includes the following sub-steps:

[0061] S11. Within the range of the delivery customer's location, randomly generate multiple cluster centers;

[0062] S12. Calculate the distance from each customer point to all cluster centers;

[0063] S13. Determine whether the minimum distance from each customer point to all cluster centers satisfies the maximum radius constraint of the nearest cluster. If yes, the customer point is added to the nearest cluster to obtain a temporary cluster, and the process jumps to step S14. If no, the process jumps to step S15.

[0064] S14. Determine whether the total weight of all customer points in the temporary cluster meets the maximum load constraint of the drone. If yes, the temporary cluster is a cluster. If not, proceed to step S15.

[0065] S15. Generate a new cluster center and jump to step S12. Continue until all customer points have corresponding clusters. Then the clustering process for delivery customer points is completed, resulting in multiple clusters.

[0066] like Figure 2 As shown, trucks loaded with all customers' goods and drones depart from distribution center N. o Starting from there, the drone travels along a predetermined route to each cluster center (drone takeoff or recovery node), requiring N... c Delivery is handled at each customer location, and each customer location has certain logistics needs. i and soft time window limit [e i ,f i After the truck arrives at the cluster center, the drone takes off and delivers goods to the customer's location based on the goods loaded by the customer within that cluster. At the same time, the truck does not need to wait for the drone to return; it can directly go to the next delivery access node to launch the drone and retrieve the drone that took off at the previous node. If the truck arrives at the node before the drone, it needs to wait there for the drone to complete battery replacement and loading of goods. This process is repeated until all delivery tasks are completed.

[0067] S2. Using the central node of the cluster as the access node for the truck and the customer points within the cluster as the access nodes for the drone (here, the cluster refers to a cluster with more than 1 customer points), plan the routes for the drone and the truck to obtain the optimal routes for the drone and the truck in the logistics delivery process.

[0068] Step S2 includes the following sub-steps:

[0069] S21. Use the distribution center as the starting access node for trucks and the central node of the cluster as the delivery access node for trucks.

[0070] S22. Use all customer points within a cluster as access nodes for a single drone takeoff.

[0071] S23. Use a genetic algorithm to process the visited nodes to obtain the optimal path for the drone and truck in the logistics delivery process.

[0072] Step S23 includes the following sub-steps:

[0073] S2301. Use natural numbers to encode drones, trucks, and access nodes;

[0074] S2302. Based on the encoded numbers, randomly generate an initial population of size Nind;

[0075] In this embodiment, assuming there are 10 customer locations, 3 clusters are obtained through the above clustering process. An initial population is generated in step S2302 as follows: Figure 3 As shown.

[0076] S2303. Inspect and remove unfeasible initial individuals;

[0077] S2304. Calculate the fitness of all individuals retained from the initial population;

[0078] Fitness value refers to the ability of an individual in a population to adapt to its environment. In this invention, fitness value is the objective function value, and the objective function is to minimize the sum of truck cost, drone cost, and time cost. Therefore, the lower the fitness value, the stronger the individual's ability to adapt to the environment. Conversely, individuals with high fitness are easily eliminated during the population evolution process.

[0079] S2305. Store high-fitness individuals in the cache unit;

[0080] S2306. Using a roulette wheel method, Nsel chromosomes are selected from the parents to obtain the initial offspring individuals; the probability of an individual being selected in step S2306 is:

[0081]

[0082] Where, p n Z represents the probability of the nth individual being selected. n Nind is the fitness value of the nth individual, and Nind is the size of the initial population.

[0083] S2307. Based on the crossover probability, select chromosomes from the initial offspring individuals for crossover operation to generate the second part of offspring individuals;

[0084] Step S2307 includes the following sub-steps:

[0085] S23071. Select chromosomes from the initial offspring individuals according to the crossover probability, decode the chromosomes, delete distribution center 0, and obtain m gene fragments. Each gene fragment is all the customer points contained in a cluster.

[0086] S23072. Randomly select two gene fragments as crossover fragments, perform crossover operation, and obtain the crossover gene fragments.

[0087] S23073. Recombine the crossover gene fragments to obtain new gene fragments;

[0088] S23074. Determine whether the total distance between the nodes to be visited by the drone in a single takeoff is within the drone's range. If not, insert the takeoff node into the end of the new gene fragment to obtain the updated gene fragment, and jump to step S23075. If yes, jump to step S23075.

[0089] S23075. Re-encode the new and updated gene fragments obtained after crossover and insert them into distribution center 0. Combine the gene fragments to form a new feasible solution, generating the second part of offspring individuals, such as... Figure 4 As shown.

[0090] This invention addresses the constraint relationship between drones and truck access nodes by improving upon the ordinary single-point crossover mechanism. It selects the client node that the drone will visit in a single takeoff as the crossover segment, which can preserve the excellent genes of the parent generation, improve crossover efficiency, and avoid generating invalid solutions.

[0091] S2308. Based on the mutation probability, select chromosomes from the second batch of offspring individuals for mutation operations to generate the third batch of offspring individuals;

[0092] Step S2308 includes the following sub-steps:

[0093] S23081. Select chromosomes from the second part of offspring individuals, decode the chromosomes, delete distribution center 0, and obtain m gene fragments. Each gene fragment is a customer point contained in a cluster.

[0094] S23082. Generate a random number P between 0 and 1 for each gene segment. κ ,κ=1,2,3,…,m,If P κ Greater than the mutation probability P m Then, the gene positions between two client points are randomly swapped to obtain the mutated gene fragment;

[0095] S23083. Recombine the mutated gene fragments, insert them into distribution center 0, and re-encode them to form a new feasible solution;

[0096] S23084. Call the judgment operator to determine whether the mutation is valid by calculating the fitness of the new feasible solution. If the fitness is less than or equal to the original fitness, it is considered valid and the mutated individual is retained. Otherwise, the individual before mutation is retained and the third part of offspring individuals is generated.

[0097] This invention adds a judgment operator to the mutation operator, which compares the individual's fitness values ​​before and after the mutation. If the fitness value after the mutation is better, it is considered a valid mutation, and the mutation result is retained; otherwise, the gene fragment before the mutation (e.g., ...) is retained. Figure 5 As shown in the figure, this improvement overcomes the shortcomings of the original algorithm in terms of low optimization efficiency, effectively improves the convergence of the algorithm, and avoids the occurrence of infeasible solutions.

[0098] S2309. Sort the individuals in the initial population in step S2304 according to fitness from low to high, select individuals with fitness from low to high (Nind-Nsel) as cache individual units, and merge the cache individual units with the third part of the offspring individuals to form a complete population of size Nind, i.e., the new offspring population.

[0099] S2310. Determine whether the number of iterations has reached the maximum number of iterations. If yes, decode the new offspring population to obtain the optimal path for the drone and the truck. If no, use the new offspring population as the parent population and jump to step S2304.

[0100] The objective function for calculating the fitness of all individuals retained in the initial population in step S2304 is:

[0101] minZ = F1 + F2 + F3

[0102] Where Z represents the individual's fitness, F1 represents the truck delivery cost, F2 represents the drone delivery cost, and F3 represents the time cost.

[0103] The formula for calculating truck delivery costs is as follows:

[0104]

[0105]

[0106] Where F1 is the truck delivery cost, c t For truck start-up costs, c T Let x be the cost per unit distance traveled by the truck, i be the i-th access node of the truck, j be the j-th access node of the truck, and x be the cost per unit distance traveled by the truck. ij Let d be the state parameter of the truck from the i-th access node to the j-th access node. ij Let N be the distance from the i-th access node to the j-th access node of the truck, and let N be the number of nodes in the road network, including the distribution center N0 and the number of customer points with delivery needs in the cluster. c Truck delivery access node N f .

[0107] The formula for calculating drone delivery costs is as follows:

[0108]

[0109]

[0110] Where F2 is the drone delivery cost, M is the number of clusters with more than 1 customer points, and c d For the startup cost of drones, c D Let N be the unit distance cost of drone flight, i be the i-th customer point in the cluster visited by the drone, j be the j-th customer point in the cluster visited by the drone, and N be the total distance cost. c Let y′ represent the number of customer points with delivery needs within the cluster. ijf Let d be the state parameters of the drone from the i-th customer point to the j-th customer point. ij Let be the distance between the i-th customer point and the j-th customer point for the drone.

[0111] The formula for calculating time cost is:

[0112] F3 = w T +w D

[0113]

[0114]

[0115] Where F3 is the time cost, w T For truck waiting time, w D Let c0 be the time penalty cost incurred due to breach of customer service hours, c0 be the unit time cost of the drone waiting for the truck, and N be the number of nodes in the road network, including the distribution center N0 and the number of customer points with delivery needs in the cluster N. c Truck delivery access node N f k T Let k be the set of nodes accessed by trucks. D Let k be the set of nodes accessed by drones, k be the number of nodes accessed by both trucks and drones, and x be the number of nodes accessed by both. ij Let d be the state parameter of the truck from the i-th access node to the j-th access node. ij Let N be the distance from the i-th access node to the j-th access node of the truck. c d represents the number of customer points with delivery needs within the cluster. ij Let y′ be the distance from the i-th access node to the j-th access node for the drone. ijf Let v be the state parameters of the drone from the i-th customer point to the j-th customer point. T v is the speed of the truck. D Let e ​​be the drone's flight speed, and e be the waiting penalty coefficient. i f is the delivery start time for the i-th customer point. i Let t be the delivery deadline for the i-th customer point. iLet f be the time it takes for the drone to reach the i-th customer point, and f be the lateness penalty coefficient.

[0116] The constraints on the objective function are:

[0117] The first constraint is that every node in the road network will be visited by a truck or drone, and only once:

[0118]

[0119] The second constraint is that drones will only begin serving customers after the trucks arrive at the delivery point:

[0120]

[0121] The third constraint prohibits truck and drone routes from forming loops:

[0122]

[0123]

[0124] The fourth constraint is that the total length of each takeoff flight path of the drone shall not exceed the maximum flight distance:

[0125]

[0126] The fifth constraint is that the total weight of all customer cargo in each cluster shall not exceed the maximum payload of the drone, meaning that each drone takeoff can serve all customers in that cluster.

[0127]

[0128] Article Six: Loading constraints for trucks must be met during delivery.

[0129]

[0130] Where, N f N is the delivery access node for trucks. c x represents the number of customer points with delivery needs within the cluster. ij Let d be the state parameter of the truck from the i-th access node to the j-th access node. ij Let y′ be the distance from the i-th access node to the j-th access node of the truck. ijf Let d be the state parameters of the drone from the i-th customer point to the j-th customer point. ij Let G be the distance from the i-th customer point to the j-th customer point, l be the maximum flight radius of the drone, and G be the distance from the i-th customer point to the j-th customer point. M Clustering customer points to form a set of all clusters, Q Dq represents the maximum payload capacity of the drone. i For customer point i's logistics demand, Q T This refers to the truck's maximum load capacity.

[0131] Example 1: A system for optimizing delivery routes using drones and trucks, comprising:

[0132] Delivery customer point clustering unit and route planning unit;

[0133] The delivery customer point clustering unit is used to cluster delivery customer points according to their location to obtain different clusters;

[0134] The path planning unit is used to plan the paths of the drone and the truck by using the central node of the cluster as the access node of the truck and the customer point within the cluster as the access node of the drone, so as to obtain the optimal path of the drone and the truck in the logistics delivery process.

[0135] The system also includes: express delivery stations, trucks, drones, and truck drivers. The trucks are specifically used to load express packages and drones, serving as a supply platform for drones. They can meet the take-off and landing requirements of drones and also have the function of replacing drone batteries and loading express packages.

[0136] experiment:

[0137] The following section further elaborates on the solution and effects of this invention by selecting a real-world rural logistics and distribution scenario.

[0138] To verify the effectiveness of the "drone + truck" joint delivery method proposed in this application, 30 villages within a town in Shaodong City, Hunan Province, were selected as customer points with delivery needs for research. The latitude and longitude coordinates of each village were obtained from Baidu Maps, projected and transformed to obtain a planar coordinate matrix, which was then normalized. It was assumed that the roads between the villages were accessible, and that each village had certain delivery needs. The drones flew along straight-line distances. To simplify the calculation, the running distances of the drones and trucks were calculated according to Euclidean distance.

[0139] Since the "drone + truck" joint delivery solution is still in the experimental operation stage and actual data is difficult to obtain, delivery time windows and delivery demand data were randomly generated based on the population and economic conditions of the study area, as shown in Table 1:

[0140] Table 1. Delivery information for express delivery points and villages.

[0141]

[0142]

[0143] Where ET represents the earliest time a customer accepts delivery service, and LT represents the latest time a customer accepts delivery service. Note: Serial number 0 represents a courier point, and 1 to 30 represent customer points with certain logistics needs; each courier point is equipped with a delivery truck and a drone, with a one-to-one correspondence between trucks and drones. See Table 2 for specific model parameters.

[0144] Table 2. Relevant parameters for trucks and drones

[0145]

[0146]

[0147] The initial population size of this invention is Nind = 100, the crossover probability is Pc = 0.9, and the mutation probability is P. m =0.2, selection probability GGAP=0.9, maximum number of iterations Max_iter=200, based on the dense distribution among villages, the cluster radius R max The initial cluster size was set to 1.5km, with P = 3 clusters. This invention uses MATLAB, specifically MATLAB 2014, to develop the algorithm. It was run on a computer with an Intel Core I5-8250U CPU, 4GB of RAM, and a 64-bit operating system. The clustering results are as follows: Figure 6 As shown in Table 3, the cluster division of customer points is shown in Table 4, and the path planning schemes for drones and trucks are shown in Table 5. Figure 7 This indicates the iterative convergence process of the method of the present invention.

[0148] Table 3 Customer Clustering Results

[0149]

[0150] The clustering results show that the 30 customer points form 11 clusters, with no customer points missing. Each cluster contains 1-4 customer points. The K-means algorithm of this invention effectively divides the customer points in the region according to the constraints of the model (maximum flight radius and maximum payload of the UAV), obtaining multiple clusters that satisfy the clustering constraints, thus verifying the effectiveness of the clustering method.

[0151] Table 4 Optimal Solution for Combined Drone and Truck Delivery

[0152]

[0153]

[0154] The optimal delivery plan shows that the drones took off 10 times, handling delivery to 29 customer locations. After takeoff, the trucks did not need to wait for the drones to return; instead, they went directly to the next cluster center to meet the drones, greatly improving the efficiency of the drone-truck joint delivery operation. Since customer number 21 was identified as a "noisy customer" during the clustering process (i.e., a cluster containing only one customer location), the truck and driver were responsible for delivering to this customer location. The final optimal route planning scheme for drones and trucks yielded a delivery cost of 162.13 yuan, a total delivery time of 1.16 hours, and a total truck travel distance of 22.27 km.

[0155] Comparative analysis

[0156] (1) Comparative Analysis of Single-Truck Delivery

[0157] To fully verify the practicality of the truck-drone joint delivery model, facing the same delivery needs, a single truck was selected to deliver to customer points. The problem was transformed into a TSP (Traffic-to-Spot) solution. To closely reflect the actual conditions of rural roads, the truck's travel distance was calculated using both straight-line distance and traffic distance. Existing research shows that the non-straight-line coefficient of rural roads ranges from 1.1 to 1.4. Therefore, the traffic distance between customer points was obtained by multiplying a coefficient matrix randomly generated within this range with the straight-line distance matrix between customer points. The final solution results for the two delivery models are shown in Table 5.

[0158] Table 5. Results of the two delivery modes

[0159]

[0160] Calculations show that the combined drone and truck delivery model increases the total cost by 20.6% and the total delivery time by 13.8% over traffic distances compared to straight-line distances. The single-truck delivery model, on the other hand, increases the total cost by 75.1% and the total delivery time by 16.5% over traffic distances. Comparing the two models, the combined drone and truck delivery model reduces the total cost by 11.2% and the total delivery time by 52.1% over straight-line distances, while reducing the total cost by 38.6% and the total delivery time by 53.2% over traffic distances.

[0161] The comparative data above shows that the speed and flexibility of drone delivery greatly improves delivery efficiency and significantly reduces delivery costs. Furthermore, the "drone + truck" joint delivery model, which is closer to actual road conditions and considers trucks operating according to traffic distance, demonstrates even more prominent economic and time-efficiency advantages. This fully illustrates that the "drone + truck" joint delivery model can effectively improve the service quality of logistics delivery in remote rural areas, allowing local residents to experience the convenience of "last mile" delivery and saving costs for express delivery companies.

[0162] (2) Comparative Analysis of Genetic Algorithms

[0163] To verify the effectiveness of the improved genetic algorithm (IGA) of this invention, a comparative analysis was conducted using a traditional genetic algorithm. The parameters of the two algorithms were unified, and the above examples were solved 10 times each. The results are shown in Table 6. The performance of IGA and GA was analyzed from three perspectives: solution results, convergence number of generations, and running time.

[0164] Table 6 Comparison of IGA and GA solution results

[0165]

[0166] As shown in Table 6, the genetic algorithm of this invention is inferior to the ordinary genetic algorithm in two out of 10 solutions. This is because the initial population has a certain degree of randomness in its encoding. However, the overall solution result of the genetic algorithm of this invention is better than that of the ordinary genetic algorithm. From the perspective of algorithm convergence, the genetic algorithm of this invention, by adding a judgment operator to improve the quality of evolved individuals and setting an elite population to retain excellent individuals, has a better convergence effect than the ordinary genetic algorithm in terms of both speed and stability. Due to the addition of the judgment operator, the computational load of the program is increased, and the program running time of the genetic algorithm of this invention is correspondingly extended.

[0167] The above analysis shows that the genetic algorithm proposed in this application has good adaptability to the model in the problem, and the overall performance of the algorithm is better than that of ordinary genetic algorithms. The effect is significant, and it can quickly obtain the optimal solution in solving the "drone + truck" joint delivery problem, thus verifying the improvement effect.

[0168] In summary, the present invention provides a method and system for optimizing the logistics delivery route of drones and trucks. Compared with the traditional "delivery personnel + truck" delivery, the joint delivery of "drones + trucks" has greatly improved the delivery efficiency and service cost. Moreover, when considering the actual road conditions, the optimization results of total cost and total time are more obvious.

Claims

1. A method for optimizing delivery routes using drones and trucks, characterized in that, Includes the following steps: S1. Based on the location of the delivery customer points, cluster the delivery customer points to obtain different clusters; including: S11. Within the range of the delivery customer's location, randomly generate multiple cluster centers; S12. Calculate the distance from each customer point to all cluster centers; S13. Determine whether the minimum distance from each customer point to all cluster centers satisfies the maximum radius constraint of the nearest cluster. If yes, the customer point is added to the nearest cluster to obtain a temporary cluster, and the process jumps to step S14. If no, the process jumps to step S15. S14. Determine whether the total weight of all customer points in the temporary cluster meets the maximum load constraint of the drone. If yes, the temporary cluster is a cluster. If not, proceed to step S15. S15. Generate a new cluster center and jump to step S12. Continue until all customer points have corresponding clusters. Then the clustering process for delivery customer points is completed, resulting in multiple clusters. S2. Using the central node of the cluster as the access node for trucks and the customer points within the cluster as the access nodes for drones, the routes of drones and trucks are planned to obtain the optimal routes for drones and trucks in the logistics delivery process.

2. The method for optimizing the logistics delivery route of drones and trucks according to claim 1, characterized in that, Step S2 includes the following sub-steps: S21. Use the distribution center as the starting access node for trucks and the central node of the cluster as the delivery access node for trucks. S22. Use all customer points within a cluster as access nodes for a single drone takeoff. S23. Use a genetic algorithm to process the visited nodes to obtain the optimal path for the drone and truck in the logistics delivery process.

3. The method for optimizing the logistics delivery route of drones and trucks according to claim 2, characterized in that, Step S23 includes the following sub-steps: S2301. Use natural numbers to encode drones, trucks, and access nodes; S2302. Based on the encoded numbers, randomly generate a number of sizes. Nind Initial population; S2303. Inspect and remove unfeasible initial individuals; S2304. Calculate the fitness of all individuals retained from the initial population; S2305. Store high-fitness individuals in the cache unit; S2306. Selecting from the parents using the roulette wheel method. Nsel One chromosome is used to obtain the initial offspring individuals; S2307. Based on the crossover probability, select chromosomes from the initial offspring individuals for crossover operation to generate the second part of offspring individuals; S2308. Based on the mutation probability, select chromosomes from the second part of offspring individuals for mutation operations to generate the third part of offspring individuals; S2309. Sort the individuals in the initial population from low to high fitness in step S2304, and select those with fitness from low to high ( Nind-Nsel Individuals are used as cached individual units, and these cached individual units are merged with the third part of the child individuals to form a group of size. Nind The complete population, i.e., the new offspring population; S2310. Determine whether the number of iterations has reached the maximum number of iterations. If yes, decode the new offspring population to obtain the optimal path for the drone and the truck. If no, use the new offspring population as the parent population and jump to step S2304.

4. The method for optimizing the logistics delivery route of drones and trucks according to claim 3, characterized in that, The objective function for calculating the fitness of all individuals retained in the initial population in step S2304 is: in, For individual fitness, For truck delivery costs, For drone delivery costs, For time cost.

5. The method for optimizing the logistics delivery route of drones and trucks according to claim 4, characterized in that, The formula for calculating the truck delivery cost is as follows: in, For truck delivery costs, For truck start-up costs, Cost per unit distance traveled by truck For the truck's first One access node, For the truck's first One access node, For the truck from the first The first access node to the truck The status parameters of each access node For the truck from the first The first access node to the truck The distance between each access node, The number of nodes in the road network, including distribution centers. The number of customer points with delivery needs in the cluster. Truck delivery access nodes .

6. The method for optimizing the logistics delivery route of drones and trucks according to claim 4, characterized in that, The formula for calculating the cost of drone delivery is as follows: in, For drone delivery costs, This represents the number of clusters in which the number of customer points is greater than 1. For the startup cost of drones, Cost per unit distance for drone flight The first cluster of drone visits One customer point, The first cluster of drone visits One customer point, This refers to the number of customer points within a cluster that have delivery needs. For drones from the first The customer clicked to the first Status parameters of each customer point For drones from the first The customer clicked to the first The distance between customer locations.

7. The method for optimizing the logistics delivery route of drones and trucks according to claim 4, characterized in that, The formula for calculating the time cost is as follows: in, For time cost, For truck waiting time, The time penalty cost incurred due to breach of customer service hours, The unit time cost of drones waiting for trucks. The number of nodes in the road network, including distribution centers. The number of customer points with delivery needs in the cluster. Truck delivery access nodes , For the set of nodes accessed by trucks, This is a set of nodes accessed by drones. The number of nodes accessed jointly by trucks and drones. For the truck from the first The first access node to the truck The status parameters of each access node For the truck from the first The first access node to the truck The distance between each access node, This refers to the number of customer points within a cluster that have delivery needs. For drones from the first The access node to the first The distance between each access node, For drones from the first The customer clicked to the first Status parameters of each customer point For the truck's speed, For the drone's flight speed, The waiting penalty coefficient, For the first Delivery start time for each customer point For the first Delivery deadline for each customer point For the drone to reach the The time for each customer point This represents the penalty coefficient for being late.

8. A system for optimizing the delivery route of unmanned aerial vehicles and trucks as described in any one of claims 1 to 7, characterized in that, include: Delivery customer point clustering unit and route planning unit; The delivery customer point clustering unit is used to cluster delivery customer points according to their location to obtain different clusters; The path planning unit is used to plan the paths of the drone and the truck by using the central node of the cluster as the access node of the truck and the customer point within the cluster as the access node of the drone, so as to obtain the optimal path of the drone and the truck in the logistics delivery process.

Images