Multi-uav-vehicle collaborative distribution system and scheduling method considering multi-access

Abstract

The application relates to the technical field of unmanned aerial vehicle air-ground collaborative logistics distribution, in particular to a multi-unmanned aerial vehicle-vehicle collaborative distribution system and a scheduling method considering multiple visits, the system comprising: at least one ground vehicle, the vehicle serving as a mobile base station, being responsible for moving on a macro path and carrying one or more unmanned aerial vehicles and packages to be distributed; and at least one unmanned aerial vehicle, the unmanned aerial vehicle being launched from the vehicle, sequentially visiting one or more customer nodes in one flight to complete a distribution task, and returning to the vehicle to replenish the packages or replace the battery after completing all the distribution tasks in the flight. The collaborative distribution system framework provided by the application allows the unmanned aerial vehicle to serve multiple customer nodes in a single flight, provides a more flexible and efficient distribution mode, can significantly reduce the return, loading and unloading and waiting time overhead of the unmanned aerial vehicle, and thus improves the operation efficiency of the whole distribution system and reduces the operation cost.

Description

Technical Field

[0001] This application relates to the field of drone air-ground collaborative logistics delivery technology, and in particular to a multi-drone-vehicle collaborative delivery system and scheduling method that considers multiple accesses. Background Technology

[0002] With the rapid and sustained development of the e-commerce industry in recent years, the demand for last-mile delivery has surged. Users are increasingly demanding faster delivery times and more precise delivery windows, making traditional logistics models increasingly unable to meet the growing last-mile delivery needs. Meanwhile, the low-altitude economy, represented by drone technology, has experienced rapid development. Drone manufacturing technology is maturing, and low-altitude airspace control is gradually opening up. Innovative "last-mile" delivery models based on drones have become a highly anticipated emerging research direction. Using drones to transport packages from last-mile delivery centers or stations to customers offers advantages such as rapid response, high efficiency, flexibility, and immunity to ground traffic congestion and terrain obstacles. However, drone delivery is currently limited by its relatively limited payload capacity and battery capacity, making it unable to independently achieve long-distance, multi-customer delivery tasks without resupply.

[0003] To overcome the aforementioned limitations of drones, truck-drone collaborative delivery has attracted increasing attention. In this model, vehicles (such as trucks) not only handle ground delivery tasks but also serve as mobile warehouses, mobile base stations, and refueling / charging stations for drones, carrying one or more drones to collaboratively execute missions. This model fully leverages the advantages of traditional truck delivery—high payload capacity and long range—combined with the rapid and flexible characteristics of low-altitude drone delivery, demonstrating significant potential in improving delivery timeliness and reducing delivery costs.

[0004] Existing research on vehicle-drone collaborative delivery has proposed a variety of frameworks. For example, the Flying-while-Delivering Traveling Salesman Problem (FSTSP) and the Traveling Salesman Problem with Drones (TSP-D) are two basic models in this field, and subsequent research has further extended to complex scenarios such as single vehicle-multiple drones (TSP-mD, mFSTSP) and multiple vehicles-multiple drones (mTSPD).

[0005] However, most existing studies impose a significant constraint: limiting drones to serving only one customer point per flight (Single-Visit). In this "single-service" mode, each time a drone departs from a vehicle, delivers a package, and then immediately returns to the vehicle for restocking or charging before performing the next mission. This mode results in frequent trips between the vehicle and the customer point, significantly increasing the time spent on launch, recovery, loading / unloading, and waiting, failing to fully utilize the drone's carrying capacity and thus limiting the overall efficiency and cost-effectiveness of the collaborative delivery system. With advancements in drone manufacturing technology, modern drones are now capable of carrying multiple packages and serving multiple customers in a single flight. This "multi-visit" capability offers greater flexibility and efficiency potential for collaborative delivery systems. By allowing drones to sequentially visit multiple customer points in a single flight, the total number of drone flights and return trips can be significantly reduced, shortening the total mission completion time (Makespan) and lowering operating costs.

[0006] However, the "multi-access" model also presents challenges: it exponentially increases the complexity of the joint planning and scheduling problem between drones and vehicles. The planner must not only decide whether a vehicle or a drone serves the customer, but also plan a complex path involving multiple customer points for each drone flight (i.e., a "sub-traveling salesman problem"), and closely coordinate it temporally and spatially with the vehicle's macro-path and the tasks of other drones. Therefore, designing an efficient scheduling method to solve this complex joint optimization problem under the "multi-access" framework has become a pressing technical challenge in this field. Summary of the Invention

[0007] This application provides a multi-drone-vehicle collaborative delivery system and scheduling method that considers multiple visits, overcoming the efficiency bottleneck caused by the "single visit" of drones in the existing drone-vehicle collaborative delivery mode.

[0008] To address the aforementioned technical problems, in a first aspect, embodiments of this application provide a multi-drone-vehicle collaborative delivery system considering multiple accesses. The system includes: at least one ground vehicle, which serves as a mobile base station, responsible for moving along a macroscopic path and carrying one or more drones and packages to be delivered; and at least one drone, which is launched from the vehicle and sequentially visits one or more customer nodes in a flight to complete delivery tasks, and returns to the vehicle for package resupply or battery replacement after completing all delivery tasks in that flight.

[0009] In some exemplary embodiments, the single flight of the drone is constrained by its maximum payload capacity and maximum battery life.

[0010] In some exemplary embodiments, the total weight of all packages delivered in a single flight does not exceed the maximum payload of the drone, and the total flight energy consumption or distance does not exceed its maximum range.

[0011] Secondly, this application provides a hierarchical task scheduling method based on a multi-UAV-vehicle collaborative delivery system considering multiple accesses as described in the above embodiments. The method includes the following steps: First, acquiring a dataset; then, performing upper-level planning and lower-level planning based on the dataset; finally, performing system execution and iteration based on the schemes output by the upper-level and lower-level planning. The system execution and iteration process includes: vehicles waiting at the customer cluster docking point, and the UAV fleet executing delivery tasks according to the optimized scheme output by the lower-level planning; when all customer points in the cluster have been served and all UAVs have returned to the vehicles, the vehicles depart for the next customer cluster docking point in the macro-path of the upper-level planning, and repeat the lower-level planning until all customer clusters have been served, and finally the vehicles return to the warehouse.

[0012] In some exemplary embodiments, acquiring the dataset includes: collecting the geographic coordinates and package weight of the customer node; collecting vehicle performance parameters and drone performance parameters respectively; wherein the vehicle performance parameters include vehicle speed and the number of drones carried by the vehicle; and the drone performance parameters include the number of drones, maximum payload, and maximum range.

[0013] In some exemplary embodiments, upper-level planning is performed based on the dataset, including: customer point clustering and vehicle macro-path generation; wherein, customer point clustering and vehicle macro-path generation includes: Step 1, based on the geographical location information of all customer nodes, using a clustering algorithm to divide all customer points into several spatially concentrated customer clusters; Step 2, using the center point of each customer cluster as the vehicle's stop point; Step 3, solving a vehicle routing problem for these stop points to determine the macro-path with the lowest cost for the vehicle to start from the warehouse, sequentially visit each customer cluster stop point, and finally return to the warehouse.

[0014] In some exemplary embodiments, lower-level planning is performed based on the dataset, including solving the intra-cluster multi-UAV multi-access task scheduling problem.

[0015] In some exemplary embodiments, solving the multi-drone multi-access task scheduling problem within a cluster includes: vehicles traveling sequentially according to the macroscopic path generated in step three; when a vehicle arrives at a stop point of a customer cluster, lower-level planning is initiated, and a genetic algorithm is used for heuristic search to solve the multi-drone multi-access delivery scheme for all customer points within the cluster.

[0016] In some exemplary embodiments, the core design of the genetic algorithm includes: first, defining a chromosome encoding scheme to represent a complete delivery scheme that allocates all customer points within a cluster to multiple drones and forms multiple "multi-access" paths; second, designing a fitness function that aims to minimize the total completion time within the cluster and penalizes solutions that violate drone payload and range constraints; and finally, iteratively performing standard genetic operations of selection, crossover, and mutation to evolve the solution population towards a high-quality optimization scheme until a termination condition is reached, thereby obtaining an approximate optimal scheduling solution within the cluster.

[0017] In some exemplary embodiments, the system execution and iteration process is carried out based on the schemes output by the upper-level and lower-level planning, including: vehicles waiting at the customer cluster docking point, and drone fleets executing delivery tasks according to the optimized scheme output by the lower-level planning. After all customer points within the cluster have been served and all drones have returned to the vehicles, the vehicles depart for the next customer cluster docking point in the macro-path of the upper-level planning, and the lower-level planning is repeated until all customer clusters have been served, and finally the vehicles return to the warehouse.

[0018] The technical solution provided in this application has at least the following advantages: This application provides a multi-drone-vehicle collaborative delivery system and scheduling method that considers multiple accesses. The system includes: at least one ground vehicle, which serves as a mobile base station, responsible for moving along a macroscopic path, and carrying one or more drones and packages to be delivered; and at least one drone, which is launched from the vehicle and sequentially visits one or more customer nodes in a flight to complete delivery tasks, and returns to the vehicle for package resupply or battery replacement after completing all delivery tasks in the flight.

[0019] The collaborative delivery system framework proposed in this application allows drones to serve multiple customer nodes in a single flight, providing a more flexible and efficient delivery model. It significantly reduces drone return, loading / unloading, and waiting time, thereby improving the overall operational efficiency and reducing operating costs. To address the exponentially increasing complexity of joint planning and scheduling under this "multi-access" framework, this application further provides a hierarchical optimization method. This method follows a "clustering first, pathfinding later" strategy, decomposing the complex joint optimization problem into two easier-to-handle sub-problems: the upper-level problem handles the clustering of customer points and the macroscopic path planning of vehicles between cluster centers; the lower-level problem handles solving the multi-access path planning for drones within each cluster and the multi-agent task allocation. For the lower-level, complex, NP-hard multi-drone multi-access path planning sub-problem, this application employs a heuristic genetic algorithm to search for high-quality drone delivery path solutions through selection, crossover, and mutation operations. Attached Figure Description

[0020] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations do not constitute a limitation on the embodiments, and unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0021] Figure 1 This is a schematic diagram of a multi-drone-vehicle scenario considering multiple accesses, provided as an embodiment of this application.

[0022] Figure 2 A flowchart of a hierarchical scheduling method provided in an embodiment of this application. Detailed Implementation

[0023] As the background technology shows, the existing "single service" model requires drones to frequently travel between vehicles and customer points, which significantly increases the time spent on drone launch, recovery, loading and unloading, and waiting, failing to fully utilize the carrying potential of drones and thus limiting the overall efficiency and cost-effectiveness of the entire collaborative delivery system.

[0024] To address the aforementioned technical problems, this application provides a multi-UAV-vehicle collaborative delivery system and scheduling method that considers multiple accesses. The system includes: at least one ground vehicle, which acts as a mobile base station, responsible for moving along a macroscopic path and carrying one or more UAVs and packages to be delivered; and at least one UAV, launched from the vehicle, which sequentially visits one or more customer nodes in a single flight to complete delivery tasks, and returns to the vehicle for package resupply or battery replacement after completing all delivery tasks in that flight. The multi-UAV-vehicle collaborative delivery system considering multiple accesses provided in this application, along with the hierarchical task scheduling method based on a "clustering first, then pathfinding" strategy, aims to solve the problem of limited overall system efficiency caused by the limitation in existing UAV-vehicle collaborative delivery frameworks, where most UAVs are restricted to serving only a single customer per flight.

[0025] The embodiments of this application will now be described in detail with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the embodiments of this application to facilitate a better understanding of the application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments.

[0026] This application provides a multi-drone-vehicle collaborative delivery system considering multiple accesses. The system includes: at least one ground vehicle, which acts as a mobile base station, responsible for moving along a macroscopic path and carrying one or more drones and packages to be delivered; and at least one drone, launched from the vehicle, which sequentially visits one or more customer nodes in a single flight to complete delivery tasks (i.e., "multi-access"), and returns to the vehicle for package resupply or battery replacement after completing all delivery tasks in that flight. A single drone flight is constrained by its maximum payload capacity and maximum battery range. The total weight of all packages delivered in a single flight does not exceed the drone's maximum payload, and the total flight energy consumption or distance (including return flight) does not exceed its maximum range.

[0027] Within the framework of the aforementioned collaborative delivery system, this application provides a hierarchical task scheduling method. This method is based on a multi-UAV-vehicle collaborative delivery system considering multiple accesses as described in the above embodiments. The method includes the following steps: First, acquiring a dataset; then, performing upper-level planning and lower-level planning based on the dataset; and finally, performing system execution and iteration based on the schemes output by the upper-level planning and lower-level planning.

[0028] This application provides a hierarchical task scheduling method. This method is based on a "Cluster-First, Route-Second" strategy, decomposing the complex joint scheduling problem into two levels of sub-problems. The specific steps of this method include: Step 1: Basic Data Collection and Algorithm Initialization: Collect the geographical coordinates and package weight of the customer node; collect the performance parameters of the vehicle and the drone, including vehicle speed and the number of drones carried by the vehicle, and drone performance parameters including the number of drones, maximum payload and maximum range. Step 2: Upper-level Planning: Customer Point Clustering and Vehicle Macro-path Generation. First, based on the geographical location information of all customer nodes, a clustering algorithm (e.g., K-Means algorithm) is used to partition all customer points into several spatially concentrated customer clusters. Then, the center point of each customer cluster (or a designated node within the cluster) is used as the vehicle's stop point. A Vehicle Routing Problem (VRP) is solved for these stop points (and warehouses) to determine the macro-path that minimizes the cost (e.g., the total travel time or distance) for the vehicle to start from the warehouse, sequentially visit each customer cluster stop point, and finally return to the warehouse.

[0029] Step 3: Lower-level planning: Intra-cluster multi-drone multi-access task scheduling. Vehicles travel sequentially according to the macro-path generated in Step 2. When a vehicle arrives at a stop point in a customer cluster, lower-level planning is initiated to solve for the multi-drone multi-access delivery scheme for all customer points within that cluster. The goal of this lower-level planning subproblem is to minimize the total time (i.e., the drone fleet completion time, Makespan) for completing delivery tasks for all customer points within the cluster, while satisfying drone payload and range constraints. For the NP-hard intra-cluster multi-drone multi-access scheduling subproblem, this application uses a Genetic Algorithm (GA) for heuristic search to solve it. The core design of the algorithm includes: First, defining a chromosome encoding method to represent a complete delivery scheme that allocates all customer points within a cluster to multiple drones and forms multiple "multi-access" paths; Second, designing a fitness function that aims to minimize the total completion time within the cluster and penalizes solutions that violate drone payload and range constraints; Finally, iteratively executing standard genetic operations such as selection, crossover, and mutation to evolve the solution population towards a high-quality optimization scheme until the termination condition is met, thereby obtaining an approximate optimal scheduling solution within the cluster.

[0030] Step 4: System Execution and Iteration: The vehicle waits at the customer cluster stop, and the drone fleet executes the delivery task according to the optimized plan output in Step 3. Once all customer points within the cluster have been served and all drones have returned to the vehicle, the vehicle departs for the next customer cluster stop on the macro-path planned in Step 2, and repeats Step 3 until all customer clusters have been served, and finally the vehicle returns to the warehouse.

[0031] The specific implementation methods of this application will be described in detail below.

[0032] This application first models a research scenario considering multi-drone-vehicle collaborative delivery with multiple accesses. A schematic diagram of the scenario is shown below. Figure 1 As shown in the diagram, green dots represent warehouses, red dots represent customer points, and blue dots represent cluster centers.

[0033] In this scenario, a depot is set up. and N customer nodes awaiting delivery Each client node With geographic coordinates and package weight .

[0034] The delivery system consists of a ground vehicle (truck) and the vehicles mounted on it. MA fleet of drones Vehicles and drones from the warehouse Departure, and after completing delivery tasks at all customer nodes, return to the warehouse. .

[0035] To ensure the real-world relevance of the research scenario and the solvability of the scheduling problem, the following necessary assumptions are made regarding the scenario: 1. Motion Constraints: Vehicles travel within the road network, and their travel distance is calculated using Manhattan distance; drones fly in the air, unrestricted by the ground road network, and their flight distance is calculated using Euclidean distance. Both vehicles and drones travel at their respective constant speeds. and Driving.

[0036] 2. Vehicle Role: The vehicle has sufficient payload capacity to carry all packages and the drone fleet, with unlimited range. The vehicle acts as a mobile depot for the drones; drones must launch from the vehicle and return to the same vehicle for resupply.

[0037] 3. Multiple access constraint for drones: The core feature of this application is that a drone is allowed to access one or more client nodes sequentially in a single flight sortie (i.e., "multiple access").

[0038] 4. Unmanned Aerial Vehicle (UAV) Capability Constraints: The number of UAV flights per sortie is limited by its maximum payload. And maximum battery life (battery capacity) The dual constraints.

[0039] Weight limit: Total weight of all packages delivered in a single shipment Not exceeding .

[0040] Range constraint: Total flight distance per sortie (from vehicle to multiple customer locations visited sequentially) The total distance (finally returning to the vehicle) Not exceeding .

[0041] 5. Task Constraints: Each customer node's package cannot be split and must be delivered in a single trip by a single drone or vehicle (if the vehicle's stop point is the customer node). Each customer node must be served exactly once.

[0042] 6. Time Ignore: To simplify the model, the time costs of loading, unloading, taking off, landing, and battery changing of the drone on the vehicle are ignored.

[0043] The optimization objective of this application is to plan the macroscopic path of the vehicle and the multi-access flight path of all drones while satisfying all the above constraints, so as to minimize the total time for the system to complete all delivery tasks.

[0044] In the aforementioned collaborative delivery scenario, this application proposes a hierarchical task scheduling method based on a "clustering first, then pathfinding" strategy. The flowchart of the method is as follows: Figure 2 As shown, the specific process is as follows: Step 1: Data collection and algorithm initialization.

[0045] In the delivery scenario, with warehouses Establish a two-dimensional coordinate system with the location as the origin. Collect all customer points. coordinates and package weight Collect and set performance parameters for the vehicle and drone fleet, including: vehicle speed. Number of drones (M), drone speed Maximum payload of drones Maximum flight time of drones .

[0046] Step 2: Upper-level planning.

[0047] Customer point clustering: The K-Means clustering algorithm is used to cluster all customers. N Each customer node is divided into positions. K A spatially concentrated customer cluster .

[0048] Number of clusters K It can be preset based on empirical values, for example, based on the expected average number of customer points within a cluster, n (e.g. ),set up After the K-Means algorithm converges iteratively, each customer point... Allocated to a single customer cluster .

[0049] Vehicle macropath generation: This step plans a macropath for the vehicle to access all customer clusters.

[0050] First, determine the docking point: for each generated customer cluster Calculate its geographical center (centroid). Select the actual client node within the cluster that is closest to its centroid. This serves as the truck stop node for the cluster.

[0051] Then, solve for the macroscopic path: K The docking points of each cluster { } and warehouse Put together, they form a collection K The Traveling Salesman Problem (TSP) with +1 node. Solve the problem using standard TSP solvers (such as OR-Tools, LKH, etc.) or heuristic algorithms, starting from where the vehicle is located. Let's set off and visit these places in turn. K Stops and eventually return. The shortest (or fastest) path This path This refers to the vehicle's macroscopic path sequence. The vehicle carries... M Deploy drones, following macroscopic paths From the warehouse Depart, heading to the first customer cluster docking point in the sequence. And moored.

[0052] Step 3: Lower-level planning.

[0053] When the vehicle arrived at the cluster stop Then, immediately initiate lower-level planning for all client nodes within the cluster (excluding docking points). (This point is directly served by the vehicle) Plan a multi-drone, multi-access delivery solution with the shortest total completion time.

[0054] Since this problem is a complex combinatorial optimization problem with NP-hard properties, this application uses a genetic algorithm (GA) to solve it.

[0055] The input to GA is: cluster The set of customer points within the area (and their coordinates and weights). M Available drones (and their) Parameters), current vehicle stop (Serving as a drone take-off and landing point).

[0056] The output of GA is: a set of multi-access flight path schemes for M UAVs.

[0057] The specific implementation of GA is as follows: (1) Encoding (Chromosome): A client-point-based encoding method is adopted. A chromosome is defined as a cluster. A permutation of all customer points within a cluster that are to be delivered (by drone). For example, if cluster There are 5 customer locations. A chromosome can be .

[0058] (2) Decoding and Fitness Functions: Decoding: A chromosome itself is merely a client point sequence; it must be converted into an actual sequence by a "decoder." M The drone executes a multi-access path scheme. The decoder employs a greedy strategy (e.g., Best-Fit or First-Fit) to minimize the completion time.

[0059] Decoder example process: a. Maintenance M 1 drone, each drone There is an "available time" (Initially all are 0).

[0060] b. Traverse the chromosome sequence .

[0061] c. For the first customer point in the sequence We will attempt to create a new flight (Route 1).

[0062] d. Continue to check the next customer point in the sequence. Check if it can be done. Add to the end of Route 1, i.e., check the new path after merging. Do the load constraints be met simultaneously? and battery life constraints .

[0063] e. If satisfied, then Join Route 1 and continue to check the next customer point. If the condition is not met, Route 1 ends (containing only...). ).

[0064] f. After a flight (e.g., Route 1) ends, calculate its total flight time T_{flight}. Select the earliest available drone (i.e., The smallest drone Assign the flight to it and update its available time. .

[0065] g. From the first unassigned customer point in the sequence (e.g. Start by repeating the cf process to build Route2.

[0066] h. Until all customer points in the chromosome sequence have been assigned to flight slots.

[0067] Fitness function: After decoding, the total completion time of this chromosome (scheme) is... Fitness value .

[0068] (3) Population initialization: randomly generated (For example, 100) different customer points are arranged (chromosomes) to form the initial population.

[0069] (4) Genetic operation (iteration): Repeat the following operations. (e.g., 500) generations: a. Selection: Tournament selection is used. Selection is done randomly from the population. k From 2 or 3 individuals, select the individual with the highest fitness as the parent. Repeat this process twice to select two parents.

[0070] b. Crossover: with Apply a crossover operator, such as order crossover (OX), to two parent chromosomes with a probability of 0.8 (e.g., 0.8) to generate two new offspring chromosomes.

[0071] c. Mutation: (The text abruptly ends here, so the translation stops as well.) Apply mutation operators, such as swap mutation (randomly swapping the positions of two client points in a chromosome sequence) to the offspring chromosomes with a probability of 0.1 (e.g., 0.1), to increase population diversity.

[0072] d. Elitism: The best-fitting individual (elite) in the current population is directly copied into the next generation of the population to ensure that the optimal solution is not lost.

[0073] (5) Termination: The maximum number of iterations is reached. Then, the algorithm terminates. The chromosome with the highest fitness recorded throughout the entire evolutionary process is the cluster. The optimal scheduling scheme.

[0074] Vehicles at the parking spot wait. M Based on the optimal solution output by the GA algorithm, each drone executes its own multi-access delivery task (which may involve multiple flights) in parallel. Vehicles simultaneously serve different stops. (If it is also a customer point).

[0075] Step 4: Iteration and Termination.

[0076] When cluster After all customer points within the area have been served and all drones have returned to the vehicle, the vehicle departs, following the macro-path. Proceed to the next customer cluster stop Upon arrival, repeat step 3 (lower-level GA solution) and execute the delivery task. When all K After serving each customer cluster, the vehicle, carrying all the drones, returned to the warehouse along the macro-path. The entire delivery task is now complete. Based on the above technical solutions, this application provides a multi-drone-vehicle collaborative delivery system and scheduling method that considers multiple accesses. The system includes: at least one ground vehicle, which serves as a mobile base station, responsible for moving along a macroscopic path and carrying one or more drones and packages to be delivered; and at least one drone, which is launched from the vehicle and sequentially visits one or more customer nodes in a flight to complete delivery tasks, and returns to the vehicle for package replenishment or battery replacement after completing all delivery tasks in the flight.

[0077] The collaborative delivery system framework proposed in this application allows drones to serve multiple customer nodes in a single flight, providing a more flexible and efficient delivery model. It significantly reduces drone return, loading / unloading, and waiting time, thereby improving the overall operational efficiency and reducing operating costs. To address the exponentially increasing complexity of joint planning and scheduling under this "multi-access" framework, this application further provides a hierarchical optimization method. This method follows a "clustering first, pathfinding later" strategy, decomposing the complex joint optimization problem into two easier-to-handle sub-problems: the upper-level problem handles the clustering of customer points and the macroscopic path planning of vehicles between cluster centers; the lower-level problem handles solving the multi-access path planning for drones within each cluster and the multi-agent task allocation. For the lower-level, complex, NP-hard multi-drone multi-access path planning sub-problem, this application employs a heuristic genetic algorithm to search for high-quality drone delivery path solutions through selection, crossover, and mutation operations.

[0078] Those skilled in the art will understand that the above-described embodiments are specific examples of implementing this application, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of this application. Any person skilled in the art can make their own modifications and alterations without departing from the spirit and scope of this application; therefore, the scope of protection of this application should be determined by the scope defined in the claims.

Claims

1. A multi-drone-vehicle collaborative delivery system considering multiple accesses, characterized in that, The system includes: At least one ground vehicle, which serves as a mobile base station, is responsible for moving along a macroscopic path and carries one or more drones and packages to be delivered; At least one drone, launched from the vehicle, sequentially visits one or more customer nodes in a flight to complete delivery tasks, and returns to the vehicle for package resupply or battery replacement after completing all delivery tasks in that flight.

2. The multi-drone-vehicle collaborative delivery system considering multiple accesses according to claim 1, characterized in that, The single flight of the drone is constrained by its maximum payload capacity and maximum battery life.

3. The multi-drone-vehicle collaborative delivery system considering multiple accesses according to claim 2, characterized in that, The total weight of all packages delivered in a single flight shall not exceed the drone’s maximum payload, and the total flight energy consumption or distance shall not exceed its maximum range.

4. A hierarchical task scheduling method, the method being implemented based on the multi-UAV-vehicle collaborative delivery system considering multiple accesses as described in any one of claims 1 to 3, characterized in that, The method includes the following steps: Obtain the dataset; Based on the dataset, perform upper-level planning and lower-level planning respectively; Based on the solutions output by the upper-level and lower-level planning, the system is executed and iterated. The system execution and iteration process includes: vehicles waiting at the customer cluster docking point, drone fleet executing delivery tasks according to the optimized plan output by the lower layer planning; when all customer points in the cluster have been served and all drones have returned to the vehicles, the vehicles set off to the next customer cluster docking point in the macro path of the upper layer planning, and repeat the lower layer planning until all customer clusters have been served, and finally the vehicles return to the warehouse.

5. The hierarchical task scheduling method according to claim 4, characterized in that, The acquisition of the dataset includes: Collect the geographic coordinates and package weight of customer nodes; Vehicle performance parameters and drone performance parameters are collected separately; the vehicle performance parameters include vehicle speed and the number of drones carried by the vehicle; the drone performance parameters include the number of drones, maximum payload, and maximum range.

6. The hierarchical task scheduling method according to claim 4, characterized in that, Based on the dataset, higher-level planning is performed, including: Customer point clustering and vehicle macro-path generation; Among them, customer point clustering and vehicle macro-path generation include: Step 1: Based on the geographical location information of all customer nodes, use a clustering algorithm to divide all customer points into several spatially concentrated customer clusters; Step 2: Designate the center point of each customer cluster as the vehicle's stopping point; Step 3: Solve a vehicle routing problem for these stops to determine the lowest-cost macro-path for a vehicle to start from the warehouse, visit each customer cluster stop in sequence, and finally return to the warehouse.

7. The hierarchical task scheduling method according to claim 6, characterized in that, Based on the dataset, lower-level planning is performed, including: Solve the problem of scheduling multiple UAVs and multiple access tasks within a cluster.

8. The hierarchical task scheduling method according to claim 7, characterized in that, Solve the intra-cluster multi-UAV multi-access task scheduling problem, including: The vehicles travel sequentially along the macroscopic path generated in step three. When a vehicle arrives at a stop in a customer cluster, the lower-level planning is initiated, and a genetic algorithm is used to perform a heuristic search to find a multi-drone, multi-access delivery solution for all customer points within that cluster.

9. The hierarchical task scheduling method according to claim 8, characterized in that, The core design of the genetic algorithm includes: First, a chromosome encoding method is defined to represent a complete delivery scheme that assigns all customer points within a cluster to multiple drones and forms multiple "multi-access" paths; Secondly, we design a fitness function that aims to minimize the total completion time within the cluster and penalizes solutions that violate the constraints on UAV payload and endurance. Finally, by iteratively performing standard genetic operations of selection, crossover, and mutation, the solution population is evolved towards a high-quality optimal solution until the termination condition is met, thereby obtaining an approximate optimal scheduling solution within the cluster.

Images