A multi-uav dynamic re-planning method for heavy-load logistics

Abstract

This invention relates to the field of dynamic replanning technology for multi-UAVs in heavy-duty logistics, and particularly to a method for dynamic replanning of multi-UAVs for heavy-duty logistics. The method includes: accessing UAV information, cargo information, and initial map information; constructing an initial route and calculating its length after field checks, coordinate unification, and obstacle avoidance processing; generating a candidate route set; then sequentially performing distance, battery power, payload, and charging time constraints to obtain a static planning model; solving the model using a fusion of simulated annealing and ant colony optimization to obtain an initial path solution; updating the map information by real-time acquisition of unknown obstacle information using millimeter-wave radar; identifying the original path obscured by obstacles to generate an old solution; and finally, dynamically updating the old solution by performing path deletion, path addition, and objective function comparison based on operator weight probabilities to obtain a replanned path solution. This invention effectively shortens the maximum time of multi-UAV missions and significantly improves the robustness and adaptability of heavy-duty logistics transportation.

Description

Technical Field

[0001] This invention relates to the field of dynamic replanning technology for multi-UAVs in heavy-duty logistics, and particularly to a dynamic replanning method for multi-UAVs in heavy-duty logistics. Background Technology

[0002] In the field of dynamic replanning technology for multi-UAVs in heavy-duty logistics, existing solutions typically acquire UAV information, cargo information, and initial map information, construct initial routes, calculate route lengths, and then model and solve based on a set of candidate routes to obtain initial path solutions. However, these solutions suffer from limitations such as difficulty in accurately distinguishing the original path state after the inclusion of unknown obstacle information, inability to support subsequent replanning processes with old solutions, and a lack of unified criteria for determining path deletion and addition. Existing methods often rely on initial map information, static planning models, or single-stage path adjustments, which can lead to the complete reprocessing of the original path after map updates. This fails to fully utilize the still-valid path order, route length, alternate landing point origin markers, and objective function results remaining in the initial path solution.

[0003] In scenarios involving multiple drones executing initial path solutions for heavy-duty logistics, after millimeter-wave radar provides real-time feedback on unknown obstacle information, it is easy to encounter problems such as mixing the original path blocked by new obstacles with the original path unaffected by new obstacles. It is also easy to encounter problems such as a lack of coordinated adjustment between drone paths with long routes and drone paths with short routes, making it difficult to meet the requirement of stable generation of new path solutions.

[0004] Regarding the joint processing of old solutions, operator weight probabilities, path deletion, path addition, objective function result comparison, current operator score update, and old solution update, existing technologies generally lack a consistent process from the updated map information to the identification of the original path blocked by new obstacles, and then to the generation of old solutions and the updating of new path solutions. It is difficult to form a continuous link of map information update, original path identification, old solution generation, path deletion, path addition, and old solution update in the dynamic replanning scenario of multi-UAVs in heavy-load logistics, resulting in the lack of a judgment basis corresponding to the old solution in the process of generating new path solutions. Summary of the Invention

[0005] To address the above problems, this invention proposes a dynamic replanning method for multiple unmanned aerial vehicles (UAVs) in heavy-duty logistics, comprising:

[0006] S100: Access drone information, cargo information and initial map information, perform field checks, coordinate unification and obstacle avoidance processing, construct initial route and calculate route length, and generate candidate route set;

[0007] S200. Based on the candidate route set, the distance, battery power, load capacity and charging time constraints are processed sequentially, and the longest time minimization model is performed to obtain the static planning model.

[0008] S300. Based on the static programming model, the initial path solution is obtained by using a combination of simulated annealing and ant colony algorithm.

[0009] S400. Based on the initial path solution, collect information on unknown obstacles and generate updated map information;

[0010] S500: Based on the updated map information, identify the original path that is obscured by obstacles and generate an old solution;

[0011] S600. Based on the operator weight probability, perform path deletion, path addition, and objective function comparison on the old solution to update and obtain the dynamic replanning path solution.

[0012] Furthermore, the process of constructing an initial route and calculating its length, and generating a candidate route set after field checking, coordinate unification, and obstacle avoidance processing includes:

[0013] The field checks include integrity checks on the drone information (drone weight, flight speed, maximum battery level, minimum battery level, real-time battery level, and payload), cargo information (cargo weight and designated location), and initial map information (takeoff point coordinates, landing point coordinates, alternate landing point coordinates, and obstacle area).

[0014] The coordinates uniformly include the coordinates of the takeoff point, landing point, alternate landing point, designated location, and obstacle area in the same map information record;

[0015] The obstacle avoidance process includes, when generating the initial route, removing connections that pass through obstacle areas and reselecting connections that do not pass through obstacle areas.

[0016] Furthermore, the process of constructing the initial route and calculating the route length includes:

[0017] The construction of the initial route involves using the coordinates of the takeoff point, the designated location, the coordinates of the landing point, and the coordinates of the alternate landing point as the endpoints of the route, and generating the initial route according to the connection relationship of areas that do not pass through obstacles;

[0018] The calculated route length includes the cumulative distance of each connection along the initial route, and the coordinates of the alternate landing point are recorded as the intermediate position of the initial route that passes through the alternate landing point.

[0019] Each record in the candidate route set contains a starting point field, an ending point field, an alternate landing point field, an obstacle area field, and a route length field.

[0020] Further, the process of sequentially processing distance, battery power, load capacity, and charging time constraints, and performing the longest time minimization model to obtain the static programming model includes:

[0021] Distance constraint processing includes determining whether the length of each initial flight path is within the maximum flight distance range of the current heavy drone based on real-time battery level, maximum battery level, minimum battery level, drone weight, and payload, and recording distance-feasible paths as distance-feasible paths;

[0022] The power constraint processing includes determining whether the power of the heavy UAV after executing the corresponding initial route is not lower than the minimum power based on the route length, real-time power, maximum power and minimum power, and recording the power-feasible path as the power-feasible path.

[0023] The load constraint processing includes determining whether a single drone's load weight is not less than the cargo weight based on the load weight and cargo weight, and recording it as a feasible path for a single drone's load weight. Alternatively, when the load weight of a single drone is less than the cargo weight, the multi-drone combined load weight determination is invoked, and combinations that satisfy a single transport are recorded as feasible paths for multi-drone combined load weight.

[0024] The charging time constraint processing includes determining whether the starting point field is consistent with the alternate landing point coordinates and the load-bearing feasible path based on the alternate landing point coordinates. If they are consistent, the time required for full charging is recorded and an alternate landing point starting point identifier is generated.

[0025] The maximum time minimization modeling process includes writing the flight path length, flight speed, alternate landing point origin identifier, and time required to fully charge each heavy UAV in the corresponding path into the same time record, and using the maximum time among multiple UAVs as the objective function result field.

[0026] The static programming model consists of the objective function result field, distance constraint field, power constraint field, load constraint field, charging time constraint field, and candidate route field.

[0027] Furthermore, the process of obtaining the initial path solution by using a fusion of simulated annealing and ant colony algorithm includes:

[0028] The method of solving the problem by combining simulated annealing and ant colony algorithm includes setting initial search parameters, constructing path set solutions, calculating the longest time, updating historical best solutions, updating pheromones, and updating transition probabilities.

[0029] The initial search parameters include the initial ant colony size, pheromone concentration, and initial temperature;

[0030] The path set solution construction includes each ant selecting a path record from the candidate route field and matching it according to the distance constraint field, power constraint field, load constraint field, and charging time constraint field to generate a path set solution containing the drone number field, cargo number field, path order field, alternate landing point starting point identifier field, route length field, time required for full charge field, and path feasibility status field.

[0031] The calculation of the longest time includes reading the path sequence field, flight path length field, and flight speed of each heavy UAV, and reading the time required to fully charge the alternate landing point starting point identifier field, and comparing them to obtain the objective function result corresponding to the solution of the path set;

[0032] The historical optimal solution update includes comparing the objective function result of the current optimal solution with the historical optimal solution. If the result is smaller, the historical optimal solution is updated, and the UAV number, cargo number, path order, flight path length, alternate landing point starting point identifier, and objective function result are recorded.

[0033] The pheromone update includes increasing the pheromone concentration of the corresponding candidate route based on the historical best solution, and adjusting the pheromone concentration based on the objective function result of the ordinary path set solution.

[0034] The transition probability update includes the transition probability constructed by generating the next round of path set solution based on the updated pheromone concentration and the feasible state of candidate routes;

[0035] The initial path solution includes the UAV number field, cargo number field, path sequence field, flight path length field, alternate landing point origin identifier field, time required for full charge field, and objective function result field.

[0036] Furthermore, the process of collecting information on unknown obstacles and generating updated map information based on the initial path solution includes:

[0037] Initiate drone execution status data collection and processing. When the heavy drone is executing the original path, it will write the current location information, drone number, cargo number, path sequence, route length, alternate landing point starting point identifier and real-time battery level into the execution path information.

[0038] Millimeter-wave radar is installed on a heavy-duty UAV to provide real-time feedback of unknown obstacle information, which includes obstacle distance, obstacle orientation, obstacle height, and a set of boundary point coordinates.

[0039] The generated updated map information includes obstacle area update processing and map information update processing;

[0040] The obstacle region update process includes incorporating the set of boundary point coordinates from the unknown obstacle information into the original obstacle region to form an updated obstacle region.

[0041] The map information update process includes replacing the obstacle areas in the initial map information with the updated obstacle areas, while retaining the coordinates of the takeoff point, landing point, and alternate landing point, and generating the updated map information.

[0042] The updated map information includes the takeoff point coordinates field, the landing point coordinates field, the alternate landing point coordinates field, the updated obstacle area field, the execution path information field, and the original path to be compared field.

[0043] Furthermore, based on the updated map information, the process of identifying the original path obscured by obstacles and generating an old solution includes:

[0044] Read the start point field, end point field, flight path length field, and path order field of the original path. Then read the set of boundary point coordinates in the updated obstacle area field. Extract the start point, end point, and flight path length of the current path segment and subsequent path segments according to the path order, and perform overlap judgment with the updated obstacle area. If an overlap is found, record the original path as the original path blocked by the new obstacle. If there is no overlap, record it as the original path unaffected by the new obstacle.

[0045] The process of generating an old solution includes: copying the path records related to the currently specified task from the initial path solution into an old solution, and marking the copied path records according to the blocked path records and the retained path records; for blocked path records, adding a mark of the original path blocked by the new obstacle to the old solution; for retained path records, adding a mark of the original path unaffected by the new obstacle to the old solution; simultaneously, writing the feasible status of the original path blocked by the new obstacle as unavailable, and writing the feasible status of the original path unaffected by the new obstacle as available; if there are original paths blocked by new obstacles in the combination of multiple UAVs corresponding to the same cargo number, writing the relevant paths under the same cargo number into the combination record together;

[0046] The old solution includes the following fields: UAV number, cargo number, path sequence, route length, alternate landing point origin identifier, time required for full charge, objective function result, original path markers blocked by new obstacles, and original path markers unaffected by new obstacles.

[0047] Furthermore, the process of performing path deletion on the old solution based on operator weight probabilities includes:

[0048] The operator weight probability includes random path deletion operator, original path deletion operator blocked by new obstacles, deletion operator for UAVs with long flight paths, random path addition operator, and path addition operator for UAVs with short flight paths; each operator corresponds to an operator score and an operator weight probability.

[0049] The path deletion includes: when the random path deletion operator is selected, selecting an original path from the available path records in the old solution and writing it into the deletion path record; when the original path blocked by a new obstacle is selected, reading the original path marker blocked by the new obstacle and writing the corresponding original path into the deletion path record; when the UAV path with a long flight path is selected for the deletion operator, reading the flight path length field and objective function result field corresponding to each UAV number, determining the UAV with a long flight path, and selecting an original path from the path order field corresponding to that UAV and writing it into the deletion path record.

[0050] Furthermore, the process of path addition and objective function comparison includes;

[0051] When the random path addition operator is selected, a path is chosen from the candidate routes that satisfy the distance constraint, power constraint, load constraint and charging time constraint and written into the new solution; when the path addition operator for the drone with a short route is selected, the route length field corresponding to each drone number in the old solution is read, the drone with the short route is determined, and the path that satisfies the above constraints is written into the path order field of the drone.

[0052] Read the objective function result fields of the new solution and the old solution. If the objective function result of the new solution is less than that of the old solution, write the new solution into the old solution update record and write the current operator into the high score record. If the objective function result of the new solution is not less than that of the old solution, write the current operator into the low score record and retain the path order field in the old solution.

[0053] Furthermore, the process of updating to obtain the dynamic replanning path solution includes:

[0054] Current operator fraction update processing and old solution update processing:

[0055] When the current operator enters a high-score record, increase the operator score and increase its operator weight probability; when the current operator enters a low-score record, decrease the operator score and decrease its operator weight probability.

[0056] During the old solution update process, if the new solution is accepted, the corresponding fields in the old solution are overwritten with the new solution; otherwise, the old solution is retained. The new path solution retains the current operator, comparison result, current operator score, and operator weight probability, which are used as input for the next replanning process.

[0057] The key innovations of this invention include:

[0058] (1) Based on the updated map information, identify the original path blocked by the new obstacle and generate the old solution. The path order, flight path length, alternate landing point starting point identifier, time required for full charge and objective function result in the initial path solution are retained in the old solution. The original path blocked by the new obstacle is distinguished and recorded from the original path not affected by the new obstacle.

[0059] (2) Based on the old solution, perform path deletion and path addition processing under operator weight probability, wherein path deletion includes randomly deleting paths, deleting original paths blocked by new obstacles, and deleting UAV paths with long flight paths, and path addition includes randomly adding paths and adding paths to UAVs with short flight paths.

[0060] (3) Based on the new solution and the old solution, the objective function result is compared, the current operator score is updated and the old solution is updated, so that the new path solution retains the correspondence between the path order, the objective function result, the current operator score and the operator weight probability.

[0061] The following are its main beneficial effects:

[0062] (1) To address the problem that the original path status is difficult to distinguish accurately after the updated map information is accessed, the original path is matched with the updated obstacle area by identifying the original path blocked by new obstacles and generating old solutions, so that the old solutions retain path status records that can be used for subsequent path deletion and path addition.

[0063] (2) To address the lack of a unified judgment basis for path deletion and path addition, path deletion and path addition processing under operator weight probability will include the original path blocked by the new obstacle, the UAV path with a long flight path and the UAV path with a short flight path into the same processing link, so that the new path solution generation process has an adjustment basis corresponding to the old solution.

[0064] (3) To address the problem of lacking a basis for judgment when replacing an old solution with a new one, the new solution and the old solution are compared under the same objective function result through objective function result comparison, current operator score update and old solution update, and the current operator score and operator weight probability are updated with the comparison result.

[0065] (4) To address the problem that existing methods are prone to reprocessing the original path after map information is updated, the original path, which is not affected by new obstacles, continues to serve as the path basis in the process of generating new path solutions by retaining the path order, flight path length, alternate landing point starting point identifier and the time required for full charging in the old solution.

[0066] (5) To address the problem that distance constraints, power constraints, load constraints and charging time constraints are difficult to continue in the replanning process of dynamic replanning of multi-UAVs in heavy-duty logistics, the candidate route set and the static planning model are called when adding a route so that the new route solution continues to correspond to the distance constraints, power constraints, load constraints and charging time constraints. Attached Figure Description

[0067] Figure 1 A flowchart illustrating a multi-UAV dynamic replanning method for heavy-duty logistics provided in this application embodiment; Figure 2 This is a schematic diagram of the initial scenario for a multi-UAV mission provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the initial path solution for multiple UAVs provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a scenario where a drone detects an unknown obstacle, provided by an embodiment of the present invention. Figure 5 This is a schematic diagram illustrating the identification of paths obstructed by obstacles, provided in an embodiment of the present invention. Figure 6 This is a schematic diagram of the replanning path after encountering an unknown obstacle, provided by an embodiment of the present invention. Figure 7 The overall flowchart of the dynamic replanning method provided in the embodiments of the present invention is shown.

[0068] A unified explanation of the meanings of the graphics and lines in the attached diagram:

[0069] 1. Node meaning: Green circles represent takeoff point A and landing point B; red circles represent designated cargo locations (a, b, c, d, e, f, g); blue circles represent alternate landing points (Ⅰ, Ⅱ, Ⅲ).

[0070] 2. Obstacle meaning: Gray shapes of different types represent known obstacle areas with different building forms; red vertical lines represent unknown obstacles detected by millimeter-wave radar in real time;

[0071] 3. Path meaning: The purple dashed line represents the preset candidate route; the blue dashed line represents the flight path of UAV #1; the purple solid line represents the flight path of UAV #2. Detailed Implementation

[0072] Example 1: Refer to Figure 1 This is a flowchart illustrating a dynamic replanning method for multi-UAVs in heavy-duty logistics provided by an embodiment of the present invention. The process may include at least steps S100-S600:

[0073] S100: Access drone information, cargo information and initial map information, perform field checks, coordinate unification and obstacle avoidance processing, construct initial route and calculate route length, and generate candidate route set;

[0074] S200. Based on the candidate route set, the distance, battery power, load capacity and charging time constraints are processed sequentially, and the longest time minimization model is performed to obtain the static planning model.

[0075] S300. Based on the static programming model, the initial path solution is obtained by using a combination of simulated annealing and ant colony algorithm.

[0076] S400. Based on the initial path solution, collect information on unknown obstacles and generate updated map information;

[0077] S500: Based on the updated map information, identify the original path that is obscured by obstacles and generate an old solution;

[0078] S600. Based on the operator weight probability, perform path deletion, path addition, and objective function comparison on the old solution to update and obtain the dynamic replanning path solution.

[0079] S100: Access drone information, cargo information and initial map information, perform field checks, coordinate unification and obstacle avoidance processing, construct initial route and calculate route length, and generate candidate route set;

[0080] When a designated task for heavy-duty logistics is established, the ground station accesses the drone information, cargo information, and initial map information of the heavy-duty drones, using these three types of information as input sources for this step. The drone information refers to the operational data of the heavy-duty drones participating in the designated task, specifically including the number of drones, drone weight, flight speed, maximum battery level, minimum battery level, real-time battery level, payload capacity, and maximum flight distance. The cargo information is the delivery data of the cargo to be transported, specifically including cargo weight, designated location, loading status, and deployment status. The initial map information is the spatial data of the designated task area, specifically including takeoff point coordinates, landing point coordinates, alternate landing point coordinates, and obstacle area. Specifically, the takeoff point coordinates record the departure position of the heavy-duty drone executing the designated task; the landing point coordinates record the recovery position of the heavy-duty drone after completing the designated task; the alternate landing point coordinates record the transit position of the heavy-duty drone for charging or waiting; and the obstacle area records areas that the heavy-duty drone must not pass through during the initial flight path construction.

[0081] Specifically, after receiving the UAV information, cargo information, and initial map information, the ground station first performs field integrity checks and coordinate consistency processing. Field integrity checks determine whether the UAV information includes UAV weight, flight speed, maximum battery level, minimum battery level, real-time battery level, and payload; whether the cargo information includes cargo weight and a specified location; and whether the initial map information includes takeoff point coordinates, landing point coordinates, alternate landing point coordinates, and obstacle areas. Coordinate consistency processing places the takeoff point coordinates, landing point coordinates, alternate landing point coordinates, specified location, and obstacle areas into the same map information record. If the takeoff point coordinates, landing point coordinates, alternate landing point coordinates, or obstacle area are missing, the ground station marks the corresponding record as unavailable and stops that record from entering the initial route construction process. If the cargo weight, payload, or real-time battery level are missing, the ground station removes the corresponding heavy UAV or cargo from the current candidate route set generation process and saves the missing field in the mission record.

[0082] Furthermore, the initial route construction process is performed by the ground station based on the initial map information. The ground station uses the takeoff point coordinates, the designated location, the landing point coordinates, and the alternate landing point coordinates as route endpoints, and generates an initial route according to the connection relationships between the route endpoints. When generating the initial route, the ground station compares each connection with an obstacle region. If a connection crosses an obstacle region, the connection is discarded, and a new connection that does not cross the obstacle region is selected. If a connection does not cross an obstacle region, the connection is recorded as a segment of the initial route. Understandably, the initial route is not the final path solution, but rather the basic route data used for subsequent distance constraints, battery constraints, payload constraints, and charging time constraints. This process transforms the location data in the map information into route data that can be used by multiple UAVs.

[0083] Furthermore, the route length calculation process is initiated after the initial route construction process is completed. The ground station calculates the distance along each connection of the initial route and sums the distances of each connection to obtain the route length corresponding to each initial route. If an initial route passes through alternate landing point coordinates, the alternate landing points passed through are recorded, and the alternate landing point coordinates are saved as the intermediate position of the initial route. If an initial route does not pass through alternate landing point coordinates, only the start point, end point, and route length are recorded. After the route length calculation process, the ground station writes the start point, end point, alternate landing points passed through, obstacle area comparison results, and route length of each initial route into the candidate route set.

[0084] Specifically, the candidate route set is the output of this step. Each record in the candidate route set corresponds to an initial route and includes a start point field, an end point field, an alternate landing point field, an obstacle area field, and a route length field. The start point field comes from the takeoff point coordinates, the designated location, or the alternate landing point coordinates. The end point field comes from the designated location, the landing point coordinates, or the alternate landing point coordinates. The alternate landing point field records whether the initial route passes through the alternate landing point coordinates. The obstacle area field records the comparison result between the initial route and the obstacle area. The route length field records the distance data of the initial route. After this step, the candidate route set is sent to S200 as input for distance constraint, power constraint, load constraint, charging time constraint, and maximum time minimization modeling processing in S200.

[0085] In the engineering implementation, a takeoff point coordinate system, a landing point coordinate system, multiple alternate landing point coordinate systems, and multiple obstacle areas are set up within the park. After two heavy-duty UAVs connect to the ground station, the ground station reads the UAV weight, flight speed, maximum battery level, minimum battery level, real-time battery level, and payload of each heavy-duty UAV, as well as the cargo weight and designated location. The ground station first connects the takeoff point coordinates to the designated location and alternate landing point coordinates, and then connects the designated location to the landing point coordinates and alternate landing point coordinates. For each connection generated, the ground station compares the connection with the obstacle areas. If a connection crosses an obstacle area, it is not included in the candidate route set. If a connection does not cross an obstacle area, the ground station calculates the route length of the connection and incorporates it into the corresponding initial route. After this processing, the records retained in the candidate route set all contain route endpoints, alternate landing point records, and route lengths. Subsequently, in step S200, the route length field is directly called for distance constraint processing.

[0086] In this step, the minimum set of core parameters includes takeoff point coordinates, landing point coordinates, alternate landing point coordinates, obstacle area, UAV weight, flight speed, maximum battery level, minimum battery level, real-time battery level, payload, cargo weight, designated location, and flight path length. When flight mode, wind speed, wind direction, temperature, humidity, and flight altitude are included as extended fields in the UAV information, the ground station binds them to the corresponding heavy-duty UAV records. These extended fields do not change the basic structure of the candidate flight path set. Understandably, after entering S200, the minimum set of core parameters participates in the modeling of distance constraints, battery constraints, payload constraints, and charging time constraints, respectively.

[0087] The technical benefits of this step can be summarized as follows: This step transforms UAV information, cargo information, and initial map information into a set of candidate routes, providing a clear input for the subsequent static planning model. The candidate route set retains the correspondence between takeoff point coordinates, landing point coordinates, alternate landing point coordinates, obstacle areas, and route lengths. This correspondence provides the foundational data for subsequent maximum time minimization modeling.

[0088] S200. Based on the candidate route set, the distance, battery power, load capacity and charging time constraints are processed sequentially, and the longest time minimization model is performed to obtain the static planning model.

[0089] This step receives the candidate route set output by S100 and uses the UAV information and cargo information from the same specified task as input to the static planning model. The candidate route set includes a start point field, an end point field, an alternate landing point field, an obstacle area field, and a route length field. The UAV information includes UAV weight, flight speed, maximum battery level, minimum battery level, real-time battery level, and payload capacity. The cargo information includes cargo weight and a specified location. The ground station writes the candidate route set, the UAV information, and the cargo information into the same task record and establishes a correspondence based on the heavy UAV number, cargo number, and initial route number. This correspondence is used to subsequently determine the feasibility of each heavy UAV on different initial routes.

[0090] Specifically, the ground station first performs distance constraint processing. This distance constraint refers to determining the flight feasibility of each initial route length based on the real-time battery level, maximum battery level, minimum battery level, UAV weight, and payload. The ground station reads the route length field from the candidate route set and the real-time battery level, maximum battery level, minimum battery level, UAV weight, and payload from the UAV information. When the ground station determines that the route length is within the maximum flight distance range of the current heavy UAV, it records the initial route as a distance-feasible path. When the ground station determines that the route length exceeds the maximum flight distance range of the current heavy UAV, it records the initial route as a distance-infeasible path and excludes this record from the static programming model. The distance-feasible path retains the start point field, end point field, alternate landing point field, and route length field, and then proceeds to the battery constraint processing.

[0091] Specifically, the ground station performs power constraint processing based on the feasible distance path. The power constraint refers to determining the power status of the heavy UAV after executing the corresponding initial route based on the route length, real-time power level, maximum power level, and minimum power level. The ground station reads the route length field from the feasible distance path and the real-time power level and minimum power level from the UAV information. When the ground station determines that the heavy UAV's power level after executing the initial route is not lower than the minimum power level, it records the feasible distance path as a power-feasible path. When the ground station determines that the heavy UAV's power level after executing the initial route is lower than the minimum power level, it records the feasible distance path as a power-infeasible path and prevents this record from entering the load constraint processing. The feasible power path retains the corresponding heavy UAV number, start point field, end point field, route length field, and power status field.

[0092] Specifically, the ground station performs load constraint processing based on the battery-feasible path. The load constraint refers to determining whether a single UAV or a combination of multiple UAVs meets the maximum load capacity for a single transport based on the payload and cargo weight. The ground station reads the cargo weight from the cargo information and the payload from the UAV information. When the ground station determines that the payload of a single heavy UAV is not less than the cargo weight, it records the corresponding battery-feasible path as a single-UAV load-feasible path. When the ground station determines that the payload of a single heavy UAV is less than the cargo weight, it invokes a multi-UAV combination load-feasibility judgment within the same specified task. The multi-UAV combination load-feasibility judgment involves matching the payload of multiple heavy UAVs with the same cargo weight and recording combinations that meet the single transport requirement as multi-UAV combination load-feasible paths. Battery-feasible paths that do not meet the load-feasibility judgment are recorded as load-infeasible paths. The single-UAV load-feasible path and the multi-UAV combination load-feasible path together form a load-feasible path.

[0093] Specifically, the ground station performs charging time constraint processing based on the feasible payload path. The charging time constraint refers to determining whether the flight path originates from the alternate landing point based on the alternate landing point coordinates and the feasible payload path, and recording the time required for full charging based on this determination. The ground station reads the origin and alternate landing point fields from the feasible payload path. When the ground station determines that the origin field matches the alternate landing point coordinates, it records the feasible payload path as the alternate landing point origin path and generates an alternate landing point origin identifier. When the ground station determines that the origin field does not match the alternate landing point coordinates, it records the feasible payload path as a non-alternate landing point origin path. For alternate landing point origin paths, the ground station reads the real-time battery level and maximum battery level from the UAV information and records the time required for full charging. For non-alternate landing point origin paths, the ground station retains the original flight path length and battery status. After this processing, the ground station obtains the charging time constraint result.

[0094] Understandably, after the ground station has completed the distance constraint, battery constraint, payload constraint, and charging time constraint, it performs a maximum time minimization modeling process. This maximum time minimization modeling process involves recording the flight path length, flight speed, alternate landing point origin identifier, and full charging time required for each heavy UAV in the corresponding path into the same time record, and using the longest time among the multiple UAVs as the objective function result field of the static programming model. The ground station then writes the distance feasible path, battery feasible path, payload feasible path, and charging time constraint results into the static programming model. The static programming model consists of the objective function result field, distance constraint field, battery constraint field, payload constraint field, charging time constraint field, and candidate flight path field.

[0095] Furthermore, the aforementioned distance constraints, power constraints, load constraints, and charging time constraints correspond to the following mathematical model for static programming of multiple UAVs for heavy-duty logistics. The objective function of this model is to minimize the maximum time for multiple UAVs to complete the task, expressed as formula (1):

[0096] in, Minimize operator; For drones; To be allocated to drones Route index; For drones The total number of routes; For drones In its first The distance flown along the route; The flight speed of the drone; As a binary decision variable, if the drone The If the route originates from an alternate landing point, the value is 1; otherwise, it is 0. For drones The time required to fully charge at the alternate landing point; This is a function to find the maximum value. For summation operations.

[0097] The objective function calculates the total time (the sum of flight time and charging time) required for each drone to execute all its routes, and takes the maximum value among all drones as the objective to be minimized, that is, to achieve "maximum time minimization".

[0098] The constraints of the above objective function include:

[0099] The distance constraint is expressed by formula (2):

[0100] in, For drones The distance coefficient is used to characterize the combined effect of payload and remaining battery power on the maximum flight range; For drones Before starting its first Real-time battery level for each flight route; For drones Minimum safe power level allowed; For drones Maximum battery capacity (full charge); For drones Its own weight; For drones In its first The weight of cargo carried on this route.

[0101] This constraint ensures that the length of each flight path does not exceed the maximum flight distance of the drone under its current payload and battery conditions.

[0102] The energy constraint is expressed by both formulas (3) and (4). Formula (3) defines the recursive relationship of real-time energy:

[0103] in, : No. The drone was performing its first Real-time battery level before the start of the flight route; For drones The power consumption coefficient is used to convert flight distance into power consumption. : No. The drone carried out its first The distance of the routes; For drones The maximum theoretical flight distance when fully charged and unloaded; For drones The initial charge; : No. The drone in its first The weight of cargo carried on this route.

[0104] This formula indicates that if the current route originates from an alternate landing point, the battery level is restored to its maximum before execution; otherwise, the battery level is deducted based on the flight distance and payload of the previous route; for the first route ( The battery level is the initial value.

[0105] Formula (4) represents the boundary constraints for electrical quantities:

[0106] Ensure that the drone's real-time battery level remains between the minimum safe level and the maximum level during each flight route.

[0107] The load constraint is expressed by formula (5):

[0108] in, To determine the maximum cargo weight that a multi-drone combination can carry in a single trip for this transportation mission, the combination can be accomplished by a single drone or multiple drones working together, while ensuring that the cargo weight does not exceed [a certain limit]. .

[0109] The charging time constraint is expressed by formula (6):

[0110] in, For drones The charging coefficient is used to calculate the time required to charge the battery from its current charge level to full charge.

[0111] This formula indicates that only when the route originates from the alternate landing point ( Charging time is generated only when the battery level is reached, and the charging time is proportional to the difference between the current battery level and the maximum battery level.

[0112] Decision variables The value constraint is expressed by formula (7):

[0113] The objective function result field records the comparison object of the longest time minimization modeling process, and the candidate route field records the initial route entering the solution process.

[0114] In this engineering implementation, two heavy-duty drones perform heavy-load logistics tasks within the park. After obtaining a set of candidate routes from S100, the ground station reads the route length of each initial route and the real-time battery level, maximum battery level, minimum battery level, drone weight, and payload of the two heavy-duty drones. The ground station first excludes initial routes whose length exceeds the maximum flight distance, and then excludes initial routes whose battery level falls below the minimum after execution. Subsequently, the ground station reads the cargo weight. If the payload of the first heavy-duty drone meets the cargo weight requirement, a feasible path for a single drone is generated. If the payload of the first heavy-duty drone does not meet the cargo weight requirement, the ground station combines the first and second heavy-duty drones as a multi-drone combination for payload determination. After completing the payload determination, the ground station reads the starting point field of the feasible payload path. If the starting point field is the coordinates of an alternate landing point, the starting point identifier and the time required for full charging are recorded. After the above processing, the ground station writes the retained path records into a static planning model and sends the static planning model to S300 as input for the fusion simulated annealing algorithm and ant colony algorithm solution processing in S300.

[0115] In this step, the minimum set of core parameters includes the candidate route set, route length, flight speed, maximum battery level, minimum battery level, real-time battery level, UAV weight, payload capacity, cargo weight, alternate landing point coordinates, and time required for full charging. When wind speed, wind direction, temperature, humidity, flight altitude, and flight mode are entered as extended fields into the UAV information, the ground station binds them to the corresponding heavy-duty UAV record and retains these extended fields in the static programming model. These extended fields do not change the basic input relationships of distance constraints, battery constraints, payload constraints, and charging time constraints. The static programming model serves as the output of this step and is invoked in S300 to generate the initial path solution.

[0116] In summary, this step transforms the candidate route set into a static programming model, integrating route length, payload, real-time battery power, and alternate landing point coordinates into the same modeling process. The static programming model preserves the correspondence between distance constraints, battery power constraints, payload constraints, and charging time constraints. This correspondence provides a fully constrained input for the S300 solution process.

[0117] S300. Based on the static programming model, the initial path solution is obtained by using a combination of simulated annealing and ant colony algorithm.

[0118] This step receives the static programming model output by S200 and uses it as input for solving the problem using a fusion of simulated annealing and ant colony optimization algorithms. The static programming model includes fields for the objective function result, distance constraints, battery power constraints, payload constraints, charging time constraints, and candidate routes. The ground station reads the initial route, route length, alternate landing point origin identifier, and full charging time from the candidate route field, and reads the feasible states from the distance constraint, battery power constraint, payload constraints, and charging time constraint fields. After reading, the ground station writes the path records that satisfy the distance, battery power, payload, and charging time constraints into the solution record. Path records that do not satisfy any constraint are not included in the path set solution construction process of this step, and the reasons for exclusion are retained in the solution record.

[0119] Specifically, the fusion of simulated annealing and ant colony optimization algorithms for solving the problem consists of initial search parameter setting, path set solution construction, maximum time calculation, historical best solution update, pheromone update, and transition probability update. The initial search parameters include the initial ant colony size, pheromone concentration, and initial temperature. The initial ant colony size is the number of ants participating in the path set solution construction during this solution process. The pheromone concentration is the intensity record of the candidate routes selected during the solution process. The initial temperature is the temperature record called by the simulated annealing algorithm when accepting changes in the path set solution. When setting the initial search parameters, the ground station writes the initial pheromone concentration of each candidate route into the candidate route field and the initial temperature into the temperature field. If the candidate route field is empty, the ground station stops this step of the solution process and records the static programming model as unsolvable.

[0120] Specifically, the ground station constructs a set of path solutions based on the initial search parameters. The set of path solutions refers to the record of path combinations assigned to multiple UAVs in a single solution. This record includes a UAV ID field, a cargo ID field, a path order field, an alternate landing point origin identifier field, a flight path length field, a time required for full charging field, and a path feasibility status field. Each ant selects a path record from the candidate flight path fields and matches it according to the distance constraint field, the battery constraint field, the payload constraint field, and the charging time constraint field. During matching, the ground station first determines whether the candidate flight path is a distance-feasible path, and then determines whether the corresponding flight path is a battery-feasible path. Afterward, the ground station determines the correspondence between the cargo weight and the payload, and determines whether the path has an alternate landing point origin identifier. After completing the above determinations, the ground station generates a set of path solutions.

[0121] Specifically, the ground station performs a maximum time calculation for each set of path solutions. During the maximum time calculation, the ground station reads the path sequence field of each heavy UAV, the flight path length field and flight speed of each path, and the time required for full charging corresponding to the alternate landing point origin identifier field. The ground station writes the flight time and full charging time of the same heavy UAV in the path sequence into a time record, and then compares the time records of multiple UAVs to obtain the objective function result corresponding to the path solution set. The objective function result is then used to determine the current optimal solution. If the objective function result of the current path solution set is less than the objective function result of an existing current optimal solution, the ground station records the current path solution set as the current optimal solution. If the objective function result of the current path solution set is not less than the objective function result of an existing current optimal solution, the ground station retains the existing current optimal solution and records the current path solution set as a normal path solution set.

[0122] Specifically, after each round of path set solution calculation, the ground station updates the historical best solution. The historical best solution is the path set solution with the smallest objective function result from the start of the solution to the current round. When the ground station determines that the objective function result of the current best solution is less than that of the historical best solution, it updates the current best solution to the historical best solution and records the UAV number, cargo number, path order, flight path length, alternate landing point origin identifier, and objective function result in the historical best solution field. If the current best solution does not change the historical best solution, the ground station retains the original historical best solution and writes the current best solution for this round into the round record. The path corresponding to the historical best solution is fixedly written into the high pheromone record during pheromone updates, while other path set solutions are entered into the ordinary pheromone record according to probability.

[0123] Specifically, the pheromone update is performed based on the historical best solution and the ordinary path set solution. The ground station first reads the path order field from the historical best solution and increases the pheromone concentration of the corresponding candidate routes. Then, the ground station reads the path order field from the ordinary path set solution and adjusts the pheromone concentration of the corresponding candidate routes according to the magnitude of the objective function result. Ordinary path set solutions with smaller objective function results are moved to higher pheromone records. Ordinary path set solutions with larger objective function results are moved to lower pheromone records. After the pheromone update is completed, the ground station updates the transition probability. The transition probability is a record of the probability that each candidate route will be selected when constructing the next round of path set solutions. The ground station generates transition probabilities based on the updated pheromone concentration and the feasibility status of the candidate routes, and writes the transition probabilities into the transition probability field.

[0124] Furthermore, the pheromone update employs a simulated annealing mechanism for probabilistic updating. Specifically, the pheromone concentration of candidate routes corresponding to historically optimal solutions will inevitably increase; while for solutions in the ordinary path set, the pheromone concentration will increase probabilistically. An update will be performed. Among other things, This represents the difference in pheromone increment between the current path set solution and the historical best solution. This represents the current simulated annealing temperature. This probability update formula allows the algorithm to accept a poor solution with a certain probability, thus avoiding premature entry into local optima. The transition probability... Based on the updated pheromone concentration Recalculate to guide the next round of ants in constructing a path set solution.

[0125] Understandably, the ground station processes the simulated annealing algorithm based on the initial temperature records. After each round of path set construction, longest time calculation, historical best solution update, pheromone update, and transition probability update, the ground station lowers the temperature record in the temperature field by one. The solution processing stops when the ground station determines that the number of iterations has reached a set number, the temperature record is below a set temperature, or the objective function result has reached a set target value. After the solution processing stops, the ground station uses the historical best solution as the initial path solution. The initial path solution includes the UAV number field, cargo number field, path order field, flight path length field, alternate landing point origin identifier field, time required for full charging field, and objective function result field. The initial path solution is sent to S400 as input for real-time millimeter-wave radar feedback of unknown obstacle information and map information updates in S400.

[0126] In this engineering implementation, two heavy-duty UAVs within the park perform multiple cargo transportation tasks. After receiving the static programming model from the S200, the ground station first reads the feasible paths based on distance, power, load, and charging time constraints. The ground station sets the initial ant colony size to thirty, the pheromone concentration for each candidate route to the same initial value, and the initial temperature to one hundred. Each ant selects a path from the candidate route field according to its transfer probability, generating a set of path solutions. The ground station calculates the longest time for each set of path solutions and records the path solution with the smallest objective function result in the current round as the current optimal solution. Subsequently, the ground station reads the path order field corresponding to the historical optimal solution and increases the pheromone concentration of the corresponding candidate route. Ordinary path solutions are entered into different pheromone records according to the objective function result. After the ground station completes multiple rounds of calculation, the historical optimal solution is written into the initial path solution. This initial path solution records the path order, corresponding cargo, corresponding alternate landing point starting point identifier, and objective function result for each heavy-duty UAV, and serves as the path basis for execution and map information updates in the next step.

[0127] In this step, the minimum set of core parameters includes the static programming model, initial ant colony size, pheromone concentration, initial temperature, path set solution, longest time, historical best solution, transition probability, and objective function result. The number of iterations, set temperature, and set target value are entered into the solution record as solution stopping fields. When wind speed, wind direction, air temperature, humidity, flight altitude, and flight mode exist as extended fields in the UAV information, the ground station binds them to the corresponding UAV ID field of the path set solution. These extended fields do not alter the basic processing chain of historical best solution updates, pheromone updates, and transition probability updates.

[0128] In summary, this step transforms the static programming model into an initial path solution while preserving the correspondence between constraint fields and path order fields. The fusion of simulated annealing and ant colony optimization algorithms ensures a continuous update process for historical optimal solutions, pheromone concentrations, and transition probabilities. This initial path solution provides an executable path foundation for the S400 map information update process.

[0129] S400. Based on the initial path solution, collect information on unknown obstacles and generate updated map information;

[0130] This step receives the initial path solution output by S300 and uses it as input for UAV execution status acquisition and map information update. The initial path solution includes a UAV ID field, a cargo ID field, a path sequence field, a flight path length field, an alternate landing point origin identifier field, a time required for full charging, and a target function result field. The ground station calls the corresponding heavy-duty UAV based on the UAV ID field and reads the original path being executed by the heavy-duty UAV based on the path sequence field. The original path is the path record already assigned to the heavy-duty UAV in the initial path solution. When the heavy-duty UAV executes the heavy-load logistics task according to the original path, the ground station initiates UAV execution status acquisition processing to obtain the execution path information.

[0131] Specifically, the UAV execution status acquisition and processing is jointly completed by the heavy-duty UAV and the ground station. When the heavy-duty UAV executes the original path, it writes its current location information, UAV number, cargo number, path sequence, flight path length, alternate landing point origin identifier, and real-time battery level into the execution path information. The current location information records the heavy-duty UAV's current position within the original path. The path sequence records the path segment currently being executed by the heavy-duty UAV. The real-time battery level records the heavy-duty UAV's battery status within the current path segment. After receiving the execution path information, the ground station matches it with the corresponding fields in the initial path solution. If the UAV number and the path sequence match, the ground station writes the execution path information into the map information update record. If the UAV number or the path sequence does not match, the ground station marks the corresponding record as abnormal and retains the original path record in the initial path solution.

[0132] Furthermore, the millimeter-wave radar is mounted on a heavy-duty UAV and executes the original path synchronously with the heavy-duty UAV. The millimeter-wave radar is used to provide real-time feedback of unknown obstacle information. This unknown obstacle information consists of obstacle data not recorded in the initial map information but collected by the millimeter-wave radar while the heavy-duty UAV executes the original path. The unknown obstacle information includes obstacle distance, obstacle orientation, obstacle height, and a set of boundary point coordinates. The obstacle distance records the distance between the heavy-duty UAV and the unknown obstacle. The obstacle orientation records the direction of the unknown obstacle relative to the heavy-duty UAV. The obstacle height records the airspace location of the unknown obstacle. The set of boundary point coordinates records the area boundary of the unknown obstacle in the map information.

[0133] Specifically, during the execution of the original path by the heavy UAV, the millimeter-wave radar collects information on unknown obstacles at a set acquisition frequency and sends this information to the ground station. Upon receiving the information, the ground station first performs a validity check. If the ground station determines that the information contains obstacle distance, obstacle orientation, obstacle height, and a set of boundary point coordinates, it writes the information into the map information update record. If the ground station determines that the obstacle distance or the set of boundary point coordinates is missing, it records the information as an anomaly and waits for the next real-time feedback from the millimeter-wave radar. If multiple consecutive real-time feedbacks are all anomaly feedback, the ground station retains the current map information and writes the corresponding execution path information of the heavy UAV into the anomaly record.

[0134] Furthermore, the ground station performs obstacle region update processing based on the execution path information and the unknown obstacle information fed back in real time by the millimeter-wave radar. This obstacle region update processing refers to writing the set of boundary point coordinates from the unknown obstacle information into the original obstacle region, forming an updated obstacle region. The ground station first reads the obstacle region from the initial map information, and then reads the set of boundary point coordinates from the unknown obstacle information. When the ground station determines that the set of boundary point coordinates is near the original path, it incorporates this set into the obstacle region. When the ground station determines that the set of boundary point coordinates is far from the original path, it still writes this set into the map information update record and marks it as an obstacle region that does not affect the current original path. After the obstacle region update processing is completed, the ground station obtains the updated obstacle region.

[0135] Specifically, after obtaining the updated obstacle area, the ground station performs map information update processing. This map information update processing involves replacing the obstacle area in the initial map information with the updated obstacle area, while retaining the takeoff point coordinates, landing point coordinates, and alternate landing point coordinates. The map information update processing also retains the start point field, end point field, and route length field from the candidate route set generated in S100. For initial routes in the candidate route set that overlap with the updated obstacle area, the ground station marks them as the original path to be compared in the map information update record. For initial routes that do not overlap with the updated obstacle area, the ground station marks them as the original path unaffected by the new obstacle in the map information update record. After the above processing is completed, the ground station generates updated map information.

[0136] Understandably, the updated map information is the output of this step. The updated map information includes takeoff point coordinates, landing point coordinates, alternate landing point coordinates, updated obstacle area field, execution path information field, and original path to be compared field. The updated obstacle area field comes from the unknown obstacle information fed back in real time by the millimeter-wave radar. The execution path information field comes from the UAV number, path sequence, and current position information collected by the heavy UAV when executing the initial path solution. The original path to be compared field records the original path that needs to be compared with the updated obstacle area later. After this step is completed, the updated map information is sent to S500 as input for S500 to identify the original path blocked by new obstacles and generate the old solution.

[0137] In this engineering embodiment, two heavy-duty UAVs within the park execute cargo transportation tasks according to the initial path solution. The first heavy-duty UAV follows the original path from the takeoff point coordinates to the designated location, while the second heavy-duty UAV follows the original path from the takeoff point coordinates via the alternate landing point coordinates to the designated location. Both heavy-duty UAVs are equipped with millimeter-wave radar. When the ground station receives the execution path information of the first heavy-duty UAV, it reads its current position information, path sequence, flight path length, and real-time battery level. Subsequently, the millimeter-wave radar of the first heavy-duty UAV feeds back information about unknown obstacles. The ground station reads the obstacle distance, obstacle orientation, obstacle height, and boundary point coordinate set from this unknown obstacle information and writes the boundary point coordinate set into the obstacle area. If the boundary point coordinate set is located near the original path of the first heavy-duty UAV, the ground station writes this original path into the "Comparable Original Path" field. If the original path of the second heavy-duty UAV does not overlap with the updated obstacle area, the ground station marks this original path as an original path unaffected by the new obstacle. After completing this process, the ground station outputs updated map information for the S500 to access.

[0138] In this step, the minimum set of core parameters includes the initial path solution, UAV number, path order, current location information, real-time unknown obstacle information from millimeter-wave radar, obstacle distance, obstacle orientation, obstacle height, boundary point coordinates, and obstacle region. When flight speed, flight altitude, flight mode, wind speed, wind direction, temperature, and humidity are entered as extended fields into the path information, the ground station binds them to the corresponding UAV number field. These extended fields do not alter the basic processing chain for writing unknown obstacle information into the obstacle region. The updated map information retains the correspondence between the initial path solution and the unknown obstacle information and is invoked in S500 to identify the original path blocked by new obstacles.

[0139] In summary, this step combines the initial path solution, the executed path information, and the real-time unknown obstacle information from the millimeter-wave radar into updated map information. The updated map information records the obstacle regions corresponding to the new obstacles and the original paths to be compared. This record provides input for the S500 to identify the original paths blocked by the new obstacles.

[0140] S500: Based on the updated map information, identify the original path that is obscured by obstacles and generate an old solution;

[0141] This step receives the updated map information output by S400 and calls the initial path solution output by S300 as the input source for identifying the original path blocked by new obstacles and generating the old solution. The updated map information includes takeoff point coordinates, landing point coordinates, alternate landing point coordinates, updated obstacle area, execution path information, and the original path to be compared. The initial path solution includes UAV number, cargo number, path order, flight path length, alternate landing point origin identifier, time required for full charge, and objective function result. The ground station reads the original path to be compared field and matches it with the path order field in the initial path solution. After matching, the ground station obtains the comparison object between the original path and the updated obstacle area.

[0142] Specifically, the identification of the original path blocked by the new obstacle refers to the ground station comparing the original path with the updated obstacle area and determining whether the original path passes through the updated obstacle area. The original path is the path record already assigned to the heavy UAV in the initial path solution. The updated obstacle area is the obstacle area record generated by the S400 based on the unknown obstacle information fed back by the millimeter-wave radar in real time. The ground station first reads the start point field, end point field, flight path length field, and path order field of the original path, and then reads the set of boundary point coordinates in the updated obstacle area field. When the ground station determines that the original path overlaps with the updated obstacle area, it records the original path as the original path blocked by the new obstacle. When the ground station determines that the original path does not overlap with the updated obstacle area, it records the original path as the original path unaffected by the new obstacle.

[0143] Furthermore, the comparison between the original path and the obstacle area is performed segment by segment by the ground station according to the path sequence. The ground station reads the current path segment and subsequent path segments of the heavy UAV from the path sequence field, and extracts the start point field, end point field, and flight path length field for each segment. Subsequently, the ground station performs an overlap judgment on each path segment with the updated obstacle area field. If the current path segment is determined to be the original path blocked by the new obstacle, the ground station writes the UAV number, cargo number, path sequence, start point field, end point field, and flight path length field of that path segment into the blocked path record. If the subsequent path segment is not determined to be the original path blocked by the new obstacle, the ground station writes that path segment into the reserved path record. The blocked path record and the reserved path record are jointly processed in the old solution generation process.

[0144] Specifically, the old solution generation process refers to the ground station copying the path records related to the currently specified task from the initial path solution into an old solution, and marking the copied path records according to the blocked path records and the retained path records. The old solution is not a path solution obtained by resolving the problem, but rather a basic path record formed after updating the map information from the initial path solution. The ground station retains the UAV number field, cargo number field, path sequence field, flight path length field, alternate landing point origin identifier field, time required for full charging field, and objective function result field in the old solution. For blocked path records, the ground station adds a marker indicating the original path is blocked by a new obstacle to the old solution. For retained path records, the ground station adds a marker indicating the original path is unaffected by new obstacles to the old solution. These markers are used for path deletion and path addition processing in S600.

[0145] Furthermore, the ground station simultaneously records the feasibility status when generating the old solution. For original paths blocked by new obstacles, the ground station records the feasibility status of the path as unavailable and sets it as the priority object for deleting original paths blocked by new obstacles. For original paths unaffected by new obstacles, the ground station records the feasibility status as available and retains its flight path length field and path order field. If only one of the multiple original paths of a heavy UAV is blocked by a new obstacle, the ground station only marks that original path and does not change the retention status of the remaining original paths of the heavy UAV. If there are original paths blocked by new obstacles in a combination of multiple UAVs corresponding to the same cargo number, the ground station writes the relevant paths under the same cargo number into the combination record for use in the path deletion and path addition processes in S600.

[0146] Understandably, in the first implementation, the ground station directly compares the original path field to be compared. This method is suitable for scenarios where the S400 has already completely written the original path record to be compared into the updated map information. The ground station only needs to read the original path field to be compared and perform overlap judgment with the updated obstacle area field. In the second implementation, the ground station reselects the comparison object based on the path order field in the initial path solution. This method is suitable for scenarios where the original path field to be compared is missing or abnormal. The ground station reads all path order fields and compares them segment by segment with the updated obstacle area field. In the third implementation, the ground station reads the current position information from the execution path information field and prioritizes comparing the path segment where the current position information is located. This method is suitable for scenarios where a heavy UAV is executing the original path. All three implementations output the original path blocked by the new obstacle and jointly enter the old solution generation process.

[0147] Specifically, during anomaly handling, if the updated map information lacks an updated obstacle area field, the ground station temporarily records the initial path solution as an old solution and marks the map information as abnormal in the old solution. If the initial path solution lacks a path order field, the ground station stops the old solution generation process and saves the abnormal field in the task record. If the updated obstacle area field exists, but the set of boundary point coordinates cannot be compared with the original path, the ground station records the corresponding original path as an unconfirmed original path and retains its UAV number, cargo number, and path order. The unconfirmed original path does not proceed to the deletion process for the original path blocked by new obstacles until the ground station receives new updated map information.

[0148] In the engineering embodiment, two heavy-duty UAVs within the park perform heavy-load logistics tasks according to the initial path solution. In the updated map information output by S400, the original path of the first heavy-duty UAV is recorded as the original path to be compared. The ground station reads the start point field, end point field, path order field, and flight path length field of the original path, and reads the set of boundary point coordinates in the updated obstacle area field. When the ground station determines that the original path passes through the updated obstacle area, it writes the original path into the blocked path record and marks it as the original path blocked by the new obstacle in the old solution. When the original path of the second heavy-duty UAV does not pass through the updated obstacle area, the ground station writes it into the reserved path record and marks it as the original path unaffected by the new obstacle in the old solution. After marking, the ground station outputs the old solution. The old solution retains the UAV number, cargo number, path order, flight path length, and objective function result of the two heavy-duty UAVs, and is used by S600 for path deletion, path addition, objective function result comparison, current operator score update, and old solution update processing under operator weight probability.

[0149] In this step, the minimum set of core parameters includes the updated map information, the updated obstacle area field, the original path to be compared field, the initial path solution, the drone number field, the cargo number field, the path order field, the flight path length field, and the objective function result field. The alternate landing point origin identifier field and the time required for full charging field are retained as constraint-related fields from the old solution and entered into S600. Current location information is included as an extension of the execution path information field in the priority comparison. These fields collectively constitute the path basis of the old solution and maintain their input relationship with the S600 path deletion and path addition processing.

[0150] Summary of the technical effects of this step: This step maps the updated map information to the initial path solution, obtaining the original path blocked by the new obstacle. The old solution retains the original path order and objective function result, and marks the path records affected by the new obstacle. This old solution provides input for path deletion and path addition under operator weight probabilities in S600.

[0151] S600. Based on the operator weight probability, perform path deletion, path addition, and objective function comparison on the old solution to update and obtain the dynamic replanning path solution.

[0152] This step receives the old solution output by S500 and uses it as input for path deletion, path addition, objective function result comparison, current operator score update, and old solution update processing. The old solution includes fields for UAV number, cargo number, path order, flight path length, alternate landing point origin identifier, time required for full charge, objective function result, original path markers blocked by new obstacles, and original path markers unaffected by new obstacles. After reading the old solution, the ground station first establishes an operator score record. This record includes operators for randomly deleting paths, deleting original paths blocked by new obstacles, deleting UAV paths with longer flight paths, randomly adding paths, and adding paths to UAVs with shorter flight paths. Each operator corresponds to an operator score and an operator weight probability. The operator weight probability is the probability record for selecting the corresponding operator in the current round.

[0153] Specifically, the ground station first performs path deletion processing. This process selects path records to be removed from the old solution. When the ground station selects a random path deletion operator, it selects an original path from the available path records in the old solution and writes this original path into the deletion path record. When the ground station selects a path deletion operator that deletes an original path blocked by a new obstacle, it reads the marker of the original path blocked by the new obstacle and writes the corresponding original path into the deletion path record. When the ground station selects a long UAV path for deletion, it reads the flight length field and objective function result field corresponding to each UAV number, determines the UAV with the long flight path, and selects an original path from the path order field corresponding to that UAV and writes it into the deletion path record. After the deletion path record is generated, the ground station writes the corresponding cargo number, origin field, destination field, flight length field, and alternate landing point origin identifier field into the path record to be added.

[0154] Specifically, after the path deletion process is completed, the ground station performs a path addition process. This path addition process reads available records from the candidate route set generated in S100 and the static programming model formed in S200, and rewrites the path record to be added into the path order. When the ground station selects the random path addition operator, it selects a path from the candidate routes that satisfy distance constraints, power constraints, payload constraints, and charging time constraints, and writes it into the new solution. When the ground station selects to add a path to a UAV with a short route, it reads the route length field corresponding to each UAV number in the old solution, determines the UAV with the short route, and writes the path that satisfies the distance constraints, power constraints, payload constraints, and charging time constraints into the path order field of that UAV. If no available path is found for the path record to be added, the ground station retains the cargo number of that record and writes the corresponding new solution into the unavailable status record.

[0155] Further, the ground station generates a new solution after path deletion and path addition processing. The new solution includes a UAV ID field, a cargo ID field, a path order field, a flight path length field, an alternate landing point origin identifier field, a time required for full charging field, and an objective function result field. This new solution is not the final output. The new solution first undergoes objective function result comparison processing. The ground station reads the objective function result fields of the new solution and the old solution and compares their sizes. If the objective function result of the new solution is less than that of the old solution, the ground station writes the new solution into the old solution update record and writes the current operator into the high-score record. If the objective function result of the new solution is not less than that of the old solution, the ground station writes the current operator into the low-score record and retains the path order field from the old solution. If the new solution has an unavailable state record, the ground station excludes the new solution from the old solution update record and writes the current operator into the low-score record.

[0156] Specifically, the current operator score update process is initiated after the objective function result comparison process is completed. The current operator is the operator selected during the current round of path deletion or path addition. When the current operator enters a high-scoring record, the ground station increases its operator score and raises its selection ratio in the operator weight probability. When the current operator enters a low-scoring record, the ground station decreases its operator score and lowers its selection ratio in the operator weight probability. Unselected operators retain their original operator scores. After the operator weight probability is updated, the ground station writes the updated operator score and operator weight probability into the next round of replanning record. This record is called in the next path deletion and path addition process.

[0157] Furthermore, the current operator score update process follows the following weight update formula:

[0158] in, Indicates the first The operator in the th... The score after the next update; This represents the score of the operator at the time of the last update; This is the historical score weighting coefficient, with a value range of [value range missing]. This is used to balance the impact of historical scores and recent average scores on the current score; For summation index variables; This represents the total number of operators; Indicates that all operators are in the th order. The sum of scores at the next update; The average score of all operators; Inside the summation symbol, it indicates the first... The fraction of each operator (and) (Synonyms)

[0159] This formula implements a moving average update of operator scores, allowing high-performing operators to receive higher scores and thus have a higher selection probability in subsequent iterations.

[0160] Meanwhile, the objective function comparison process adopts The greedy strategy updates the old solution: if the objective function value of the new solution is less than that of the old solution, the new solution is accepted and the old solution is updated simultaneously, while the current operator is given a higher score (its operator score is increased); otherwise, the old solution is not updated. The probability of rejecting the current solution update is [not specified], and the current operator is given a low score (its operator score is reduced). Wherein, The greedy factor has a value range of 1. This represents the probability threshold for directly accepting a better solution. This strategy strikes a balance between utilizing historical best solutions and exploring new solutions, effectively improving the adaptability of dynamic reprogramming.

[0161] Specifically, the old solution update process is performed based on the objective function result comparison process. When the ground station accepts the new solution, it overwrites the corresponding fields in the old solution with the UAV number, cargo number, path order, flight path length, alternate landing point origin identifier, time required for full charge, and objective function result fields from the new solution, and uses the overwritten path record as the new path solution. When the ground station does not accept the new solution, it retains the UAV number, cargo number, path order, flight path length, alternate landing point origin identifier, time required for full charge, and objective function result fields from the old solution, and uses the retained path record as the new path solution. The new path solution retains the current operator, comparison result, current operator score, and operator weight probability for use in the next map information update process or the next path deletion and path addition process.

[0162] In an engineering implementation, two heavy-duty drones perform heavy-load logistics tasks within the park. In the old solution output by the S500, one original path of the first heavy-duty drone is marked as blocked by a new obstacle, while the path record of the second heavy-duty drone is marked as unaffected by the new obstacle. After reading the old solution, the ground station selects to delete the original path blocked by the new obstacle based on the operator weight probability, and writes the blocked original path of the first heavy-duty drone into the deleted path record. Subsequently, the ground station reads the path record to be added and determines that the second heavy-duty drone belongs to the drone with a shorter route based on the route length field. The ground station selects to add a path to the drone with a shorter route operator, reads paths that satisfy distance constraints, power constraints, load constraints, and charging time constraints from the candidate route set, and writes them into the path order field of the second heavy-duty drone. After the new solution is generated, the ground station compares the objective function results of the new solution and the old solution. If the objective function result of the new solution is smaller, the ground station updates the old solution and writes the high-score record for the two current operators in this round. If the objective function result of the new solution does not decrease, the ground station retains the old solution and writes a low score record to the current operator in this round.

[0163] In this step, the minimum set of core parameters includes the old solution, the original path markers blocked by new obstacles, the original path markers unaffected by new obstacles, operator scores, operator weight probabilities, path deletion, path addition, the new solution, the objective function result, the current high operator score, the current low operator score, and the new path solution. The candidate route set, distance constraints, battery constraints, payload constraints, and charging time constraints serve as inputs for the path addition process. Current location information, real-time battery level, and alternate landing point origin identifier are bound as extended fields to the corresponding UAV ID field and participate in the feasibility status determination during the path addition process.

[0164] In summary, this step transforms the old solution into a new path solution, preserving the label differences between the original paths blocked by new obstacles and those unaffected by them. Operator weight probabilities, current operator scores, and objective function results are continuously updated. The new path solution retains the path order, objective function results, and operator weight probabilities for subsequent replanning processes.

[0165] To further illustrate the practical application of the method described in this invention, a specific task scenario is provided below as an example. Please refer to the accompanying drawings: Figure 2 This is a schematic diagram of the initial scenario for a multi-UAV mission provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the initial path solution for multiple UAVs provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a scenario where a drone detects an unknown obstacle, provided by an embodiment of the present invention. Figure 5This is a schematic diagram illustrating the identification of paths obstructed by obstacles, provided in an embodiment of the present invention. Figure 6 This is a schematic diagram of the replanning path after encountering an unknown obstacle, provided by an embodiment of the present invention. Figure 7 The overall flowchart of the dynamic replanning method provided in the embodiments of the present invention is shown.

[0166] exist Figure 2 In the task scenario shown, two heavy drones (numbered 1# and 2#) depart from starting point A (corresponding to the takeoff point coordinates in the initial map information). They need to transport cargo sequentially through designated locations a, b, c, d, e, and g, finally reaching the destination B (corresponding to the landing point coordinates) for retrieval. The gray shapes in the diagram represent known obstacle areas (corresponding to the obstacle areas in the initial map information), and I, II, and III represent alternate landing point coordinates, used for charging and extending the drone's range when its battery is low. The purple lines show some preset candidate routes (i.e., the initial route constructed in step S100), but not all feasible paths; supplementary routes from path points to alternate landing points are not fully displayed.

[0167] By executing steps S100 to S300, which involves accessing drone information, cargo information, and initial map information, and generating a candidate route set through field checks, coordinate unification, and obstacle avoidance processing, a static planning model is obtained by sequentially applying distance, battery power, payload, and charging time constraints. Finally, the model is solved using a fusion of simulated annealing and ant colony optimization algorithms, yielding the following results: Figure 3 The initial path solution is shown. Figure 3 The image shows the initial paths assigned to two drones: for example, the initial path for drone #1 is... The path of UAV #2 is another route that avoids obstacles. Each path records the route length, the starting point of the alternate landing point, the time required to fully charge, and the objective function result (i.e., the longest mission completion time among multiple UAVs).

[0168] During the execution of step S400, assuming that when UAV 1# is passing through position f, its onboard millimeter-wave radar provides real-time feedback on an unknown obstacle (this obstacle is located between positions d and f, marked with a red vertical line, such as...). Figure 4 (As shown).

[0169] The unknown obstacle is Figure 4 The path marked with a red vertical line is the original path of drone #1. This caused the entire flight path segment to be completely blocked. Figure 5 The results show how the system identifies all original path segments obscured by the obstacle, based on updated map information.

[0170] After receiving information about the unknown obstacle, the ground station merges its boundary point coordinates into the existing obstacle area to generate updated map information. This unknown obstacle disrupts the original path... The flight path segment is blocked and needs to be replanned. Step S500 identifies the original path blocked by the new obstacle (i.e., the aforementioned flight path segment of UAV 1#) and marks it as unavailable, while marking the unaffected paths (such as the complete path of UAV 2# and other flight path segments) as available, generating the old solution.

[0171] Subsequently, step S600 performs path deletion and path addition operations on the old solution based on operator weight probabilities. The ground station prioritizes selecting the "delete the original path blocked by the new obstacle" operator, and... After the blocked line segment is deleted, a feasible path to bypass the new obstacle is reselected from the candidate route set based on the operator "add path to drone with short flight path" or "randomly add path" (e.g., ...). (etc.), generating a new solution. By comparing the objective function (the new solution's longest time is less than the old solution), the new solution is accepted and the operator score is updated, ultimately yielding, as shown below. Figure 6 The new path solution shown, for example, the new path of UAV #1 is adjusted as follows: They successfully bypassed the unknown obstacles.

[0172] Figure 7 This is an overall flowchart of the method of the present invention, which is related to... Figure 1 The steps S100 to S600 shown correspond to each other. Figure 7 In this process, S1 corresponds to steps S100 to S300 (establishing a static programming model and solving for the initial path), S2 corresponds to step S400 (updating map information in real time), and S3 is the task completion judgment. If the task is not completed, it proceeds to S4 (i.e., dynamic reprogramming in steps S500 to S600). This cycle repeats until all drones reach the landing point and all cargo is correctly delivered to the designated location. In case of an emergency, the drone sends a warning signal and its current location information to the ground station and lands to await further instructions.

[0173] The above examples illustrate that the present invention can effectively handle path changes caused by unknown obstacles in heavy-duty logistics scenarios, and can quickly generate new path solutions through dynamic replanning, thereby significantly improving the robustness and adaptability of multi-UAV transportation.

Claims

1. A dynamic replanning method for multiple unmanned aerial vehicles (UAVs) in heavy-duty logistics, characterized in that, include: S100: Access drone information, cargo information and initial map information, perform field checks, coordinate unification and obstacle avoidance processing, construct initial route and calculate route length, and generate candidate route set; S200. Based on the candidate route set, the distance, battery power, load capacity and charging time constraints are processed sequentially, and the longest time minimization model is performed to obtain the static planning model. S300. Based on the static programming model, the initial path solution is obtained by using a combination of simulated annealing and ant colony algorithm. S400. Based on the initial path solution, collect information on unknown obstacles and generate updated map information; S500: Based on the updated map information, identify the original path that is obscured by obstacles and generate an old solution; S600. Based on the operator weight probability, perform path deletion, path addition, and objective function comparison on the old solution to update and obtain the dynamic replanning path solution.

2. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 1, characterized in that, The process of constructing an initial route and calculating its length, and generating a set of candidate routes after field checking, coordinate unification and obstacle avoidance processing, includes: The field checks include integrity checks on the drone information (drone weight, flight speed, maximum battery level, minimum battery level, real-time battery level, and payload), cargo information (cargo weight and designated location), and initial map information (takeoff point coordinates, landing point coordinates, alternate landing point coordinates, and obstacle area). The coordinates uniformly include the coordinates of the takeoff point, landing point, alternate landing point, designated location, and obstacle area in the same map information record; The obstacle avoidance process includes, when generating the initial route, removing connections that pass through obstacle areas and reselecting connections that do not pass through obstacle areas.

3. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 2, characterized in that, The process of constructing the initial route and calculating the route length includes: The construction of the initial route involves using the coordinates of the takeoff point, the designated location, the coordinates of the landing point, and the coordinates of the alternate landing point as the endpoints of the route, and generating the initial route according to the connection relationship of areas that do not pass through obstacles; The calculated route length includes the cumulative distance of each connection along the initial route, and the coordinates of the alternate landing point are recorded as the intermediate position of the initial route that passes through the alternate landing point. Each record in the candidate route set contains a starting point field, an ending point field, an alternate landing point field, an obstacle area field, and a route length field.

4. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 3, characterized in that, The process of sequentially handling distance, battery power, load capacity, and charging time constraints, and then performing the longest time minimization model to obtain the static programming model includes: Distance constraint processing includes determining whether the length of each initial flight path is within the maximum flight distance range of the current heavy drone based on real-time battery level, maximum battery level, minimum battery level, drone weight, and payload, and recording distance-feasible paths as distance-feasible paths; The power constraint processing includes determining whether the power of the heavy UAV after executing the corresponding initial route is not lower than the minimum power based on the route length, real-time power, maximum power, and minimum power, and recording the power-feasible path as a power-feasible path; the load constraint processing includes determining whether the load of a single UAV is not less than the load weight based on the load weight and cargo weight, and recording it as a single UAV load-feasible path, or calling the multi-UAV combined load judgment when the load weight of a single UAV is less than the load weight, and recording the combination that satisfies the single transport as a multi-UAV combined load-feasible path. The charging time constraint processing includes determining whether the starting point field is consistent with the alternate landing point coordinates and the load-bearing feasible path based on the alternate landing point coordinates. If they are consistent, the time required for full charging is recorded and an alternate landing point starting point identifier is generated. The maximum time minimization modeling process includes writing the flight path length, flight speed, alternate landing point origin identifier, and time required to fully charge each heavy UAV in the corresponding path into the same time record, and using the maximum time among multiple UAVs as the objective function result field. The static programming model consists of the objective function result field, distance constraint field, power constraint field, load constraint field, charging time constraint field, and candidate route field.

5. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 4, characterized in that, The process of obtaining the initial path solution by using a fusion of simulated annealing and ant colony algorithm includes: The method of solving the problem by combining simulated annealing and ant colony algorithm includes setting initial search parameters, constructing path set solutions, calculating the longest time, updating historical best solutions, updating pheromones, and updating transition probabilities. The initial search parameters include the initial ant colony size, pheromone concentration, and initial temperature; The path set solution construction includes each ant selecting a path record from the candidate route field and matching it according to the distance constraint field, power constraint field, load constraint field, and charging time constraint field to generate a path set solution containing the drone number field, cargo number field, path order field, alternate landing point starting point identifier field, route length field, time required for full charge field, and path feasibility status field. The calculation of the longest time includes reading the path sequence field, flight path length field, and flight speed of each heavy UAV, and reading the time required to fully charge the alternate landing point starting point identifier field, and comparing them to obtain the objective function result corresponding to the solution of the path set; The historical optimal solution update includes comparing the objective function result of the current optimal solution with the historical optimal solution. If the result is smaller, the historical optimal solution is updated, and the UAV number, cargo number, path order, flight path length, alternate landing point starting point identifier, and objective function result are recorded. The pheromone update includes increasing the pheromone concentration of the corresponding candidate route based on the historical best solution, and adjusting the pheromone concentration based on the objective function result of the ordinary path set solution. The transition probability update includes the transition probability constructed by generating the next round of path set solution based on the updated pheromone concentration and the feasible state of candidate routes; The initial path solution includes the UAV number field, cargo number field, path sequence field, flight path length field, alternate landing point origin identifier field, time required for full charge field, and objective function result field.

6. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 5, characterized in that, The process of collecting information on unknown obstacles and generating updated map information based on the initial path solution includes: Initiate drone execution status data collection and processing. When the heavy drone is executing the original path, it will write the current location information, drone number, cargo number, path sequence, route length, alternate landing point starting point identifier and real-time battery level into the execution path information. Millimeter-wave radar is installed on a heavy-duty UAV to provide real-time feedback of unknown obstacle information, which includes obstacle distance, obstacle orientation, obstacle height, and a set of boundary point coordinates. The generated updated map information includes obstacle area update processing and map information update processing; The obstacle region update process includes incorporating the set of boundary point coordinates from the unknown obstacle information into the original obstacle region to form an updated obstacle region. The map information update process includes replacing the obstacle areas in the initial map information with the updated obstacle areas, while retaining the coordinates of the takeoff point, landing point, and alternate landing point, and generating the updated map information. The updated map information includes the takeoff point coordinates field, the landing point coordinates field, the alternate landing point coordinates field, the updated obstacle area field, the execution path information field, and the original path to be compared field.

7. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 6, characterized in that, Based on the updated map information, the process of identifying the original path obscured by obstacles and generating an old solution includes: Read the start point field, end point field, flight path length field, and path order field of the original path. Then read the set of boundary point coordinates in the updated obstacle area field. Extract the start point, end point, and flight path length of the current path segment and subsequent path segments according to the path order, and perform overlap judgment with the updated obstacle area. If an overlap is found, record the original path as the original path blocked by the new obstacle. If there is no overlap, record it as the original path unaffected by the new obstacle. The process of generating an old solution includes: copying the path records related to the currently specified task from the initial path solution into an old solution, and marking the copied path records according to the blocked path records and the retained path records; for blocked path records, adding a mark of the original path blocked by the new obstacle to the old solution; for retained path records, adding a mark of the original path unaffected by the new obstacle to the old solution; simultaneously, writing the feasible status of the original path blocked by the new obstacle as unavailable, and writing the feasible status of the original path unaffected by the new obstacle as available; if there are original paths blocked by new obstacles in the combination of multiple UAVs corresponding to the same cargo number, writing the relevant paths under the same cargo number into the combination record together; The old solution includes the following fields: UAV number, cargo number, path sequence, route length, alternate landing point origin identifier, time required for full charge, objective function result, original path markers blocked by new obstacles, and original path markers unaffected by new obstacles.

8. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 7, characterized in that, The process of performing path deletion on the old solution based on operator weight probability includes: The operator weight probability includes random path deletion operator, original path deletion operator blocked by new obstacles, deletion operator for UAVs with long flight paths, random path addition operator, and path addition operator for UAVs with short flight paths; each operator corresponds to an operator score and an operator weight probability. The path deletion includes: when the random path deletion operator is selected, selecting an original path from the available path records in the old solution and writing it into the deletion path record; when the original path blocked by a new obstacle is selected, reading the original path marker blocked by the new obstacle and writing the corresponding original path into the deletion path record. When selecting a drone path with a long flight path for deletion, the flight path length field and objective function result field corresponding to each drone number are read to determine the drone with the long flight path, and an original path is selected from the path order field corresponding to that drone and written into the deletion path record.

9. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 8, characterized in that, The process of path addition and target function comparison includes: When the random path addition operator is selected, a path is chosen from the candidate routes that satisfy the distance constraint, power constraint, load constraint and charging time constraint and written into the new solution; When selecting to add a path to the drone operator with a short flight path, read the flight path length field corresponding to each drone number in the old solution, determine the drone with the short flight path, and write the path that satisfies the above constraints into the path order field of the drone. Read the objective function result fields of the new solution and the old solution. If the objective function result of the new solution is less than that of the old solution, write the new solution into the old solution update record and write the current operator into the high score record. If the objective function result of the new solution is not less than that of the old solution, write the current operator into the low score record and retain the path order field in the old solution.

10. The multi-UAV dynamic replanning method for heavy-duty logistics according to claim 9, characterized in that, The process of updating to obtain the dynamic replanning path solution includes: Current operator fraction update processing and old solution update processing: When the current operator enters a high-score record, increase the operator score and increase its operator weight probability; when the current operator enters a low-score record, decrease the operator score and decrease its operator weight probability. During the old solution update process, if the new solution is accepted, the corresponding fields in the old solution are overwritten with the new solution; otherwise, the old solution is retained. The new path solution retains the current operator, comparison result, current operator score, and operator weight probability, which are used as input for the next replanning process.

Images