Multi-unmanned aerial vehicle distribution cooperative task planning method

Abstract

The application provides a multi-unmanned aerial vehicle (UAV) distribution cooperative task planning method, and relates to the technical fields of UAV distribution task planning and intelligent optimization. The method obtains the coordinates, task weight and task priority information of a distribution center and customer nodes, constructs a distance matrix and a heuristic information matrix, filters feasible candidate customer nodes based on a dynamic residual range budget model, and constructs a multi-UAV cooperative distribution path through sharing of an unvisited customer node set, generation of a transfer probability, and a single-step competition mechanism. Further, ranking type global pheromone updating, adaptive adjustment of a volatilization factor, and pheromone pruning are performed according to a comprehensive evaluation score, and multi-round extension planning is performed when all customer nodes are not covered in a single round of distribution. The method can improve the accuracy of candidate path feasibility determination, improve the coordination of multi-UAV task allocation, and improve the stability and engineering applicability of distribution task planning results.

Description

Technical Field

[0001] This invention belongs to the field of multi-drone delivery task planning and intelligent optimization technology, and in particular relates to a multi-drone delivery collaborative task planning method. Background Technology

[0002] In multi-drone delivery processes, drones typically depart from the same distribution center, travel to multiple delivery points to complete their tasks, and then return to the distribution center. This process requires consideration not only of the path lengths between delivery points but also of the drone's maximum payload, flight range, and the priority differences between different delivery tasks. Existing technologies typically employ shortest path planning methods, vehicle path planning methods, and intelligent optimization methods such as ant colony optimization, genetic algorithms, and particle swarm optimization to solve this type of problem. While these methods can accomplish delivery path searching to some extent, they still have significant limitations when applied to real-world multi-drone delivery scenarios.

[0003] On the one hand, many existing task planning methods treat the maximum range of drones as a fixed value. However, in actual delivery processes, the payload of drones changes as the delivery task is executed, and this change in payload further affects the drone's flyable range. If a fixed range model is still used, it is easy to cause inaccurate judgment of path feasibility, which in turn leads to inconsistencies between the planning results and the actual execution conditions.

[0004] On the other hand, in multi-drone collaborative delivery, the task allocation and route selection among different drones are mutually influential. Without an effective coordination mechanism, problems can easily arise such as some drones being overloaded with tasks, others underloaded, unreasonable local route selection, and overall low delivery efficiency. These problems are particularly pronounced when there are a large number of delivery points with a dispersed spatial distribution.

[0005] Furthermore, in actual delivery tasks, different delivery points typically have different levels of urgency or service priority. Many existing methods primarily use the shortest total path length or lowest total cost as optimization objectives, rarely taking into account the priority completion requirements of high-priority tasks. Therefore, they are difficult to meet the application scenarios that require both delivery timeliness and task importance.

[0006] For multi-UAV mission planning methods based on ant colony algorithms, existing technologies generally suffer from a relatively simplistic pheromone update method. For example, updating the global pheromone based solely on a single optimal solution can easily lead to premature convergence, sensitivity to random initial conditions, and significant fluctuations in results, thus affecting the stability and reliability of the overall mission planning outcome.

[0007] Meanwhile, when a single-round delivery cannot cover all delivery points, existing technologies often lack a unified and clear follow-up expansion mechanism, making it difficult to directly form a continuously executable multi-round delivery solution. This limits the application of existing methods in engineering projects.

[0008] The above analysis shows that existing multi-UAV delivery task planning technologies still have shortcomings in areas such as load and range coupling constraint modeling, multi-UAV collaborative allocation, task priority consideration, and solution stability. Therefore, a multi-UAV collaborative delivery task planning method is needed to improve the feasibility, stability, and delivery execution efficiency of the task planning results. Summary of the Invention

[0009] The purpose of this invention is to provide a multi-UAV delivery collaborative task planning method, which enables extended planning of remaining tasks when a single round of delivery cannot cover all delivery points, thereby improving the feasibility, stability, and delivery execution efficiency of the delivery task planning results. This addresses the following problems in existing technologies: insufficient dynamic processing of load-range coupling constraints, leading to biased feasibility judgments of candidate delivery paths; lack of a multi-UAV dynamic collaborative allocation mechanism based on a shared set of unvisited tasks, easily resulting in uncoordinated task allocation; insufficient joint evaluation of task priority and arrival time, making it difficult to balance task execution cost with timely completion of high-priority tasks; and problems with the ant colony optimization-based solution process, such as rapid pheromone concentration, large result fluctuations, and insufficient solution stability.

[0010] To solve the above-mentioned technical problems, the specific technical solution of the present invention is as follows:

[0011] A multi-drone delivery collaborative task planning method, characterized in that it is applied to a delivery system consisting of one distribution center, drones and This is a delivery scenario consisting of several customer nodes, where the node number of the delivery center is denoted as 0, and its coordinates are shown as follows: The node IDs of all client nodes constitute the client node ID set. For the first There are 1 client node, whose coordinates are: The task weight is recorded as Task priority is recorded as The maximum payload of each drone is The maximum equivalent range when unloaded is The maximum effective range when fully loaded is The method includes the following steps:

[0012] Step S1: Obtain delivery task data and construct basic planning information: Obtain node information of the delivery center and all customer nodes, calculate the distance between customer nodes based on the coordinate information of each customer node, construct a distance matrix, construct a heuristic information matrix based on the distance matrix, and construct and initialize the pheromone matrix.

[0013] Step S2: Screening feasible candidate customer node set based on dynamic remaining range budget: Establish dynamic remaining range budget model, combine the currently constructed path, candidate customer nodes and the returning distribution center to form a candidate delivery path record, and perform segment-by-segment remaining range budget update; if the final remaining range budget obtained after completing the budget update of all segments of the candidate delivery path is greater than or equal to zero and the load constraint is less than the maximum load, then the candidate delivery path is determined to be feasible, and the candidate customer node is added to the set of feasible candidate customer nodes of the UAV;

[0014] Step S3: Construct ant colonies and generate candidate access targets based on transition probabilities: In each iteration, construct multiple ant colonies. Each ant colony is used to independently generate a set of multi-drone delivery paths. Based on the paths currently constructed by each drone in the ant colony, determine the set of visited customer nodes of the ant colony. Based on the shared set of unvisited customer nodes, determine the set of feasible candidate customer nodes for each drone. Calculate the transition probabilities based on the pheromone matrix and the heuristic information matrix to generate candidate access targets for each drone.

[0015] Step S4: Update the multi-drone cooperative path based on the single-step competition mechanism to obtain the ant colony solution: For each drone that generates candidate access behavior, calculate its competition cost; select the drone with the minimum competition cost as the winner and execute the current step of delivery; after the winning drone completes the access edge selection, perform local pheromone update on the access edge.

[0016] Step S5: Perform a comprehensive evaluation of the ant colony solutions and execute global pheromone updates, adaptive volatile factor adjustments, and pheromone pruning: Calculate the comprehensive objective value and constraint penalties for each ant colony solution to obtain the comprehensive evaluation score; select the ant colony with the best comprehensive evaluation score. Each optimal solution participates in the ranking-based global pheromone update; the volatile factor is adaptively adjusted, and the upper and lower bounds of the pheromone matrix are pruned.

[0017] Step S6: Output the single-round delivery result and perform multi-round extended planning, and output the final delivery plan: Determine whether the termination condition is met; when the single-round delivery planning result does not cover all customer nodes, extract the remaining customer node set and reconstruct the subproblem, and repeat steps S1 to S5; repeat the extended planning process until all customer nodes have completed delivery, and output the final delivery plan.

[0018] Further, in step S1, the formula for calculating the elements of the heuristic information matrix is:

[0019]

[0020] in, Indicates the first The first customer node and the first Heuristic information between client nodes This represents the maximum distance in the distance matrix. Indicates the first Task priority weights for each client node This represents the sum of the priority weights of all client node tasks; when or At that time, corresponding heuristic information Set to 0.

[0021] Furthermore, in step S2, the update formula for the remaining range budget is:

[0022]

[0023] in, This indicates the estimated remaining distance before the start of the current flight segment. This indicates the budget for the remaining distance after completing the current flight segment. Indicates the current flight distance. This indicates the remaining weight of the drone to be delivered at the start of the current flight segment. Indicates the maximum load capacity. Indicates the maximum equivalent range when unloaded. This indicates the maximum equivalent range when fully loaded.

[0024] Furthermore, in step S3, let the first... The ant colony in the 1st The set of feasible candidate client nodes obtained after the drone traverses unvisited client nodes according to step S2 and completes the judgment of load constraints and remaining range budget is as follows: Under the constraint of sharing the set of unvisited customer nodes, the set of feasible candidate customer nodes for the UAV to participate in the transfer probability calculation is defined as follows:

[0025]

[0026] in, Indicates the first The set of visited client nodes of an ant colony; the transition probability satisfies:

[0027]

[0028] in, For the first The current set of feasible candidate customer nodes for drones Candidate customer nodes in the process, Represents client node To candidate customer nodes pheromone intensity, Represents client node To candidate customer nodes Heuristic information, Represents client node To candidate customer nodes pheromone intensity, Represents client node To candidate customer nodes Heuristic information, Represents pheromone factor, Represents the heuristic factor.

[0029] Further, in step S4, the formula for calculating the competition cost is:

[0030]

[0031] in, Indicates the first The ant colony in the 1st The competitive cost of a drone in the current path construction step; Indicates the first The drone's current location is compared to the candidate access targets generated for the drone in step S3. distance, Indicates the candidate access target Distance to the distribution center Indicates candidate delivery routes The remaining range budget is obtained after updating the remaining range budget for each segment. Indicates the first The ant colony in the 1st The total load on the corresponding path after a drone adds candidate customer nodes to the current path. and The competition cost weighting coefficient, To prevent tiny positive numbers with a denominator of zero.

[0032] Further, in step S4, the local pheromone update formula is:

[0033]

[0034] in, Indicates time Side pheromone concentration, Indicates time side The concentration of pheromones on the surface Indicates the local pheromone volatile factor. This indicates the initial pheromone concentration.

[0035] Further, in step S5, the formula for calculating the comprehensive target value is:

[0036]

[0037] in, Indicates the first The overall objective value of the ant colony solution. This represents the ant colony index. Indicates the drone index; This represents the length of the complete mission execution path of the u-th drone in the k-th ant colony. Indicates the baseline path length. This represents the priority-weighted arrival cost of the u-th drone in the k-th ant colony. Indicates the number of client nodes. This represents the sum of the task priority weights of all client nodes. and These represent the weight coefficients for the path length item and the priority-weighted timeliness item, respectively.

[0038] The formula for calculating the comprehensive evaluation score is as follows:

[0039]

[0040] in, Indicates the first The overall evaluation score of the ant colony solution. Indicates the first The total range penalty for an ant colony to resolve Indicates the first The total load penalty for solving an ant colony. and These represent the range penalty weight and the load penalty weight, respectively.

[0041] Furthermore, in step S5, the ant colony solutions are sorted from smallest to largest according to their comprehensive evaluation scores, and the top scores are selected. The optimal solutions are selected as candidate solutions to participate in the global pheromone update, and their ranking weights are calculated:

[0042]

[0043] in, Indicates the first The ranking weight of each optimal solution;

[0044] Let the first The set of path edges contained in each optimal solution is The preferred solution lies on the edge. pheromone increment Defined as:

[0045]

[0046] in, Indicates the initial pheromone concentration. Indicates the first The comprehensive evaluation score of each optimal solution;

[0047] side The pheromones on the surface are updated globally in a ranked manner as follows:

[0048]

[0049] in, Indicates time Side pheromone concentration, Indicates time side The concentration of pheromones on the surface Indicates the first Global pheromone evaporation factor at the next iteration.

[0050] Further, in step S5, the adaptive adjustment formula for the volatile factor is:

[0051]

[0052]

[0053] in, Indicates the first The local pheromone evaporation factor used in this iteration Indicates the first The global pheromone evaporation factor used in this iteration Indicates the first The local pheromone evaporation factor used in this iteration Indicates the first The global pheromone evaporation factor used in this iteration and These represent the lower bounds of the local volatile factor and the global volatile factor, respectively. and These represent the attenuation coefficients of the local volatile factor and the global volatile factor, respectively.

[0054] After completing the global pheromone update, the pheromone matrix is ​​pruned using upper and lower bounds. The pruning formula is as follows:

[0055]

[0056] in, This represents the edges after ranking-based global pheromone updates and upper and lower bound clipping. pheromone concentration, and These represent the lower and upper bounds of the permissible pheromone concentration, respectively.

[0057] Furthermore, in the multi-round extended task planning process based on the sub-problem of reconstructing the remaining unfinished customer nodes, the proportion of remaining unfinished customer nodes is used to represent the degree of unfinished tasks in the current round, and this proportion is used as an additional penalty item in the calculation of the comprehensive evaluation score during the multi-round extended task planning process.

[0058]

[0059]

[0060] in, Indicates the first The percentage of remaining uncompleted client nodes in an ant colony. Indicates the first The number of customer nodes that have not completed delivery in the ant colony solution. Indicates the total number of customer nodes. Indicates the first The overall evaluation score of the ant colony solution. Indicates the first The overall objective value of the ant colony solution. Indicates the first The total range penalty for an ant colony to resolve Indicates the first The total load penalty for solving an ant colony. and These represent the range penalty weight and the load penalty weight, respectively. This indicates the penalty weight for unfulfilled customer nodes.

[0061] Compared with the prior art, the present invention has the following beneficial technical effects:

[0062] (1) By establishing a dynamic remaining range budget model and directly introducing it into the feasibility screening process of candidate delivery routes, this invention can more accurately reflect the differences in range consumption caused by changes in payload during the delivery process of UAVs, thereby improving the accuracy of route feasibility determination in the mission planning process.

[0063] (2) By constructing a multi-UAV collaborative path construction mechanism based on a shared set of unvisited client nodes and combining it with a single-step competitive decision-making method, this invention can improve the coordination of task allocation among multiple UAVs and reduce local conflicts and unreasonable allocation in the task planning process.

[0064] (3) By incorporating the path length index and the task priority weighted arrival index into the comprehensive objective function, this invention can control the delivery path cost while taking into account the completion time of high priority tasks, thereby improving the adaptability of task planning results to task timeliness requirements.

[0065] (4) By adopting a ranking-based global pheromone update, adaptive adjustment of volatile factors, and pheromone upper and lower bound pruning mechanism, this invention can reduce the excessive dominance of a single optimal solution on the pheromone evolution process, improve the situation of pheromone concentration too quickly, and thus improve the stability of the task planning solution process.

[0066] (5) When a single round of delivery cannot cover all customer nodes, the present invention can output a continuously executable delivery plan by performing multi-round extended planning on the remaining tasks, thereby improving the engineering applicability of the method in complex delivery scenarios. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a general schematic diagram illustrating the application of the present invention in a multi-drone delivery scenario, used to explain the relationship between the delivery center, customer nodes, drones, and task execution.

[0069] Figure 2 This is a technical roadmap for the multi-drone delivery collaborative task planning method of the present invention, used to illustrate the overall process of the present invention from task input, constraint modeling, collaborative solution to multi-round extended output.

[0070] Figure 3 This is a schematic diagram illustrating the determination of feasible candidate paths under the dynamic remaining range budget constraint of the present invention, used to explain the feasibility screening process based on the segmented budget update of candidate paths.

[0071] Figure 4 This is a block diagram of the pheromone evolution mechanism of the present invention, used to illustrate the relationship between local pheromone updates, ranking-based global pheromone updates, adaptive adjustment of volatile factors, and pheromone upper and lower bound pruning.

[0072] Figure 5 This is a schematic diagram of the multi-round extended delivery of the present invention, used to illustrate the process of reconstructing sub-problems for the remaining customer nodes and continuing task planning when a single round of delivery cannot cover all customer nodes.

[0073] Figure 6 The diagram shows the results of multi-round delivery task planning in an embodiment of the present invention.

[0074] Figure 7 The diagram shows the multi-round piecewise convergence results of the embodiment, illustrating the multi-round extended solution convergence process of the method of the present invention in the embodiment. Detailed Implementation

[0075] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] like Figure 1 As shown, this invention proposes a multi-drone delivery collaborative task planning method, applicable to a system consisting of one delivery center, drones and This is a delivery scenario consisting of several customer nodes, where the node number of the delivery center is denoted as 0, and its coordinates are shown as follows: The node IDs of all client nodes constitute the client node ID set. For the first There are 1 client node, whose coordinates are: The task weight is recorded as Task priority is recorded as The maximum payload of each drone is The maximum equivalent range when unloaded is The maximum effective range when fully loaded is ;like Figure 2 As shown, the method includes the following steps:

[0077] Step S1: Obtain delivery task data and build basic planning information.

[0078] Step S11: Obtain delivery task data. The delivery task data includes node information of the delivery center and all customer nodes. The node information includes node number, two-dimensional plane coordinates, task weight, and task priority.

[0079] Furthermore, a customer node refers to a delivery point that needs to perform a delivery task.

[0080] Step S12: Calculate the distance between customer nodes based on the coordinate information of each customer node, and construct a distance matrix.

[0081] Specifically, the distance between client nodes refers to the Euclidean distance between client nodes, and its calculation formula is as follows:

[0082]

[0083] in, Indicates the first The first customer node and the first Euclidean distance between client nodes Indicates the first The coordinates of each customer node. Indicates the first The coordinates of each client node are given. The distances between all client nodes form a distance matrix, represented as follows: .

[0084] Step S13: Construct a heuristic information matrix based on the distance matrix to characterize the distance factors and task priority factors of customer nodes.

[0085] Specifically, the heuristic information between client nodes is calculated using the Euclidean distance between them, and the calculation formula is as follows:

[0086]

[0087] in, Indicates the first The first customer node and the first Heuristic information between client nodes This represents the maximum distance in the distance matrix. Indicates the first Task priority weights for each client node This represents the sum of the priority weights of all client node tasks; when or At that time, corresponding heuristic information Set to 0.

[0088] The heuristic information among all client nodes constitutes a heuristic information matrix, represented as follows: .

[0089] Step S14: Construct and initialize the pheromone matrix.

[0090] Specifically, construct the initial pheromone matrix. ,in and ; Represents a node With nodes The current pheromone intensity of the path edges between them. All elements of the pheromone matrix. All are initialized to the initial pheromone value. .

[0091] Step S2: Filter the set of feasible candidate customer nodes based on the dynamic remaining range budget.

[0092] like Figure 3 As shown, based on the equivalent maximum range of the UAV without load. The maximum effective range when fully loaded is Maximum load is Establish a dynamic remaining range budget model; assuming the first... The current path constructed by the drone is , This indicates the estimated remaining distance before the start of the current flight segment. This indicates the budget for the remaining distance after completing the current flight segment. Indicates the current flight distance. This indicates the remaining weight to be delivered on the drone at the start of the current flight segment. The remaining range budget is then updated using the following formula:

[0093]

[0094] Before updating the budget segment by segment for any candidate delivery route, the initial remaining flight budget corresponding to that route is calculated. Set as the equivalent maximum range of the UAV when unloaded For candidate customer nodes The currently constructed path Candidate customer nodes The combination of return warehouses and distribution centers forms a candidate delivery route record. For the candidate delivery routes Perform segment-by-segment remaining range budget updates. Among them, This represents the intermediate remaining range budget obtained after the budget update for a single flight segment; Indicates candidate delivery routes The final remaining flight budget is obtained after all segments of the flight have been updated sequentially. If... and , Indicates the first A drone will transport candidate customer nodes The total load of the corresponding path after adding the currently constructed path is used to determine the candidate client node. The corresponding candidate delivery route is feasible, so the corresponding candidate customer node is added to the first... The set of feasible candidate customer nodes for each drone is determined; otherwise, the candidate delivery route is deemed infeasible, and the corresponding candidate customer node is not added to the list. A set of feasible candidate customer nodes for the drone.

[0095] Step S3: Construct an ant colony and generate candidate access targets based on the transition probability.

[0096] In the ant colony algorithm for the current single-round delivery problem, In the next iteration, multiple ant colonies are constructed, each ant colony is used to independently generate a set of multi-drone delivery routes, and the th iteration... An ant colony is denoted as And each ant colony contains A drone; set up the first The ant colony in the 1st The current path constructed by the drone is Its current position is recorded as Based on the paths already constructed by each drone within the ant colony, determine the set of client nodes already visited by the ant colony. .

[0097] Let the first The ant colony in the 1st The set of feasible candidate client nodes obtained after the drone traverses unvisited client nodes according to step S2 and completes the judgment of load constraints and remaining range budget is as follows: Under the constraint of sharing the set of unvisited customer nodes, the set of feasible candidate customer nodes for the UAV to participate in the transfer probability calculation is defined as follows:

[0098]

[0099] in, Indicates the first The current set of feasible candidate customer nodes for the drone.

[0100] The shared set of unvisited client nodes is used to ensure that different drones within the same ant colony do not access client nodes repeatedly.

[0101] Let the first The node number of the current client node where the drone is located is The node number of the candidate client node is Then, based on the current pheromone matrix and the heuristic information matrix, the first... Select candidate customer nodes using drones transition probability The transition probability is calculated as follows:

[0102]

[0103] in, For the first The current set of feasible candidate customer nodes for drones Candidate customer nodes in the process, Represents client node To candidate customer nodes pheromone intensity, Represents client node To candidate customer nodes Heuristic information, Represents pheromone factor, Represents the heuristic factor; based on the transition probability, in the set A customer node is selected according to the roulette wheel method, as the first... The ant colony in the 1st Candidate visit targets for drones This is used for subsequent competitive decision-making.

[0104] Step S4: Update the multi-drone cooperative path based on the single-step competition mechanism to obtain the ant colony solution.

[0105] For the ant colony, let the ant colony be the first The candidate access targets generated by the drone in step S3 are Its current constructed path is The current location is For each drone that generates a candidate access behavior, a corresponding candidate delivery path is constructed. The candidate delivery path is constructed in the following way:

[0106]

[0107] in, Indicate candidate delivery routes; This indicates a path concatenation operation that appends nodes sequentially to the current path. For the candidate delivery paths... Perform the segment-by-segment remaining range budget update in step S2 to obtain the first segment. The remaining range budget obtained after updating the remaining range budget segment by segment for each segment of the candidate delivery route completed by the drone And calculate the first The competitive cost of a drone in the current path construction step The cost of competition is calculated as follows:

[0108]

[0109] in, Indicates the first The ant colony in the 1st The competitive cost of a drone in the current path construction step; Indicates the first The drone's current location is compared to the candidate access targets generated for the drone in step S3. distance, Indicates the candidate access target Distance to the distribution center Indicates candidate delivery routes The remaining range budget is obtained after updating the remaining range budget for each segment. Indicates the first The ant colony in the 1st The total load on the corresponding path after a drone adds candidate customer nodes to the current path. and The competition cost weighting coefficient, To prevent tiny positive numbers with a denominator of zero;

[0110] In the Within each ant colony, the competition cost of all drones that have generated candidate access behaviors is compared, and the drone with the lowest competition cost is selected. and its candidate access targets As the winner;

[0111] Only the winning drone is allowed Perform the current step of task allocation and path update operation. The path update operation includes: adding candidate access targets... Add to the end of the winning drone's current path, making Update the current location of the winning drone to Increase the cumulative path length by the distance of the currently visited edge. Increase the total load of the corresponding path by the client node. The task weight; at the same time, candidate access targets From the Remove the drone from the shared set of unvisited client nodes of the ant colony, and the remaining drones remain in their current positions during the current path construction step.

[0112] Among them, the currently accessed edge Indicates the winning drone From the current node Fly to candidate visit target The resulting path edge; after the currently visited edge is determined, a local pheromone update is performed on the path edge, and the local pheromone update formula is:

[0113]

[0114] in, Indicates time Side pheromone concentration, Indicates time side The concentration of pheromones on the surface Indicates the local pheromone volatile factor. This indicates the initial pheromone concentration.

[0115] Repeat steps S3 to S4 until the shared set of unaccessed client nodes is empty, or when none of the drones in the ant colony have a feasible candidate access target, then determine the first... Each ant colony completes a multi-UAV collaborative path construction and obtains the corresponding ant colony solution.

[0116] Step S5: Perform a comprehensive evaluation of the ant colony's solution and execute pheromone learning and updating.

[0117] In this step, the set of multi-UAV paths obtained by an ant colony after completing a multi-UAV cooperative path construction is called an ant colony solution, which serves as a candidate solution for comprehensive evaluation and pheromone updating.

[0118] Step S51: Calculate the comprehensive target value of the ant colony solution based on path length and priority-weighted arrival cost.

[0119] When the The multi-drone path set obtained after an ant colony completes one multi-drone cooperative path construction is:

[0120]

[0121] in, Indicates the first Multiple drone path sets in an ant colony Indicates the first The ant colony in the 1st The complete delivery route for the drone.

[0122] For the Imagine a drone delivering goods, and let the sequence of customer nodes it visits along its complete delivery route be denoted as . , Indicates the first Each customer node Indicates the first The ant colony in the 1st The number of customer nodes visited by the drone during the complete delivery route; the corresponding complete route length is denoted as... ,set up This indicates that the drone departed from the distribution center and arrived at the destination. Customer Nodes The cumulative flight distance over time is then denoted as the priority-weighted arrival cost of the UAV:

[0123]

[0124] in, Indicates the first The first ant colony Priority-weighted arrival cost of drones Represents client node The corresponding task priority weight.

[0125] set up Let represent the reference path length obtained by the greedy algorithm or the nearest neighbor algorithm, used to normalize the path length term. Then, the ... The overall objective value of ant colony decomposition Defined as:

[0126]

[0127] in, and These represent the weighting coefficients for the path length item and the priority-weighted timeliness item, respectively. Indicates the first The ant colony in the 1st The complete path length of the drone.

[0128] Step S52: Based on the comprehensive objective value of the ant colony solution, calculate the total range penalty and total load penalty of the ant colony solution.

[0129] Specifically, based on the comprehensive target value, the first... The constraint penalty value of each ant colony solution, which includes distance penalty and load penalty.

[0130] Step S521: Calculate the drone's range penalty. The drone's range penalty is calculated as follows:

[0131]

[0132] in, Indicates the first The first ant colony The range penalty of drones Indicates the first The first ant colony The remaining flight budget is obtained by updating the remaining flight budget segment by segment after the drone completes the entire flight path.

[0133] Step S522: Calculate the payload penalty of the drone. The payload penalty of the drone is calculated as follows:

[0134]

[0135] in, Indicates the first The first ant colony Weight penalty for drones Indicates the first The first ant colony The total load of the drone along its corresponding path.

[0136] Step S523: Calculate the total range penalty and total payload penalty for ant colony solutions. The total range penalty and total payload penalty for ant colony solutions are calculated as follows:

[0137]

[0138] in, , They represent the first The total range penalty and total load penalty for each ant colony solution.

[0139] Step S53: Calculate the comprehensive evaluation score of the ant colony solution based on the total range penalty and total payload penalty.

[0140] Specifically, the comprehensive evaluation score of ant colony resolution is calculated as follows:

[0141]

[0142] in, Indicates the first The overall evaluation score of the ant colony solution. and These represent the range penalty weight and the load penalty weight, respectively.

[0143] Step S54: Sort the ant colony solutions from smallest to largest based on their comprehensive evaluation scores, and select the top... The optimal solutions are selected as candidate solutions to participate in the global pheromone update, and their ranking weights are calculated.

[0144] Specifically, the ranking weight of the optimal solution is calculated as follows:

[0145]

[0146] in, Indicates the first The ranking weight of each optimal solution.

[0147] Step S55: Perform a ranking-based global pheromone update based on the ranking weights and comprehensive evaluation scores corresponding to each optimal solution.

[0148] like Figure 4 As shown, the pheromone evolution process includes local pheromone updates, ranking-based global pheromone updates, adaptive adjustment of volatile factors, and pheromone upper and lower bound pruning.

[0149] Let the first The path edge sets corresponding to the optimal solutions are: If the side Belongs to path edge set Then the preferred solution is opposite to the edge Generate pheromone increment; if edge Not part of the path edge set If the optimal solution does not produce an pheromone increment on that edge, then the optimal solution will not produce an pheromone increment on that edge. For the th Let there be a optimal solution, and its comprehensive evaluation score be . Then the optimal solution is on the edge pheromone increment Defined as:

[0150]

[0151] in, Indicates the initial pheromone concentration. Indicates the first The comprehensive evaluation score of each optimal solution.

[0152] Based on the previous The pheromone increment and ranking weight of the optimal solution, for the edge Perform a ranked global pheromone update:

[0153]

[0154] in, Indicates the first Global pheromone evaporation factor at the next iteration.

[0155] Step S56: Update the local pheromone evaporation factor and the global pheromone evaporation factor.

[0156] Specifically, to achieve adaptive adjustment of the volatile factor, the local pheromone volatile factor and the global pheromone volatile factor are updated according to the following formulas:

[0157]

[0158]

[0159] in, Indicates the first The local pheromone evaporation factor used in this iteration Indicates the first The global pheromone evaporation factor used in this iteration Indicates the first The local pheromone evaporation factor used in this iteration Indicates the first The global pheromone evaporation factor used in this iteration and These represent the lower bounds of the local volatile factor and the global volatile factor, respectively. and Let represent the decay coefficients of the local volatile factor and the global volatile factor, respectively; after completing the global pheromone update, the pheromone matrix is ​​pruned by upper and lower bounds, and the pruning formula is as follows:

[0160]

[0161] in, This represents the edges after ranking-based global pheromone updates and upper and lower bound clipping. pheromone concentration, and These represent the lower and upper bounds of the permissible pheromone concentration, respectively.

[0162] Step S6: Output the single-round delivery results and perform multi-round expansion planning.

[0163] In this invention, single-wheel delivery planning refers to All drones depart from the distribution center and form a delivery path from the distribution center back to the distribution center in this round of planning; steps S1 to S5 are used to solve the current single-round delivery planning. After the comprehensive evaluation and global pheromone update in step S5 of each iteration, it is determined whether the current solution process meets the preset termination condition; the preset termination condition includes reaching the maximum number of iterations. Optionally, when early termination judgment is enabled, if consecutive Within the iteration window formed by the iterations, the current optimal comprehensive evaluation score The improvement amount is less than the preset convergence threshold If the conditions for early termination are met, the current single-round delivery planning result is output. If the conditions for termination are not met, the process returns to step S3 and continues to execute the multi-drone collaborative path construction, competitive decision-making, solution evaluation, and pheromone update process.

[0164] like Figure 5 As shown, when the single-round delivery planning result does not cover all customer nodes, the set of completed customer nodes is determined based on the customer nodes that have been visited in the current single-round delivery planning result, and the customer node set is... Unvisited client nodes are considered as the set of remaining client nodes. Based on the remaining set of customer nodes The remaining task sub-problems are reconstructed with the delivery center nodes to obtain the corresponding sub-distance matrix, sub-heuristic information matrix, and remaining task data. Then, while keeping the number of drones and constraint parameters unchanged, steps S1 to S5 are repeated for the sub-problems to obtain the next round of delivery planning results. The above process of repeatedly solving the remaining task sub-problems with the remaining customer node set as input constitutes the multi-round extended planning process of this invention. The above process of extracting remaining tasks, reconstructing sub-problems, and replanning is repeated until all customer nodes complete delivery or the preset maximum number of rounds is reached. Finally, the delivery results of each round are summarized in round order, outputting the multi-round task planning scheme, the delivery paths of each drone in each round, the arrival order of each customer node, and the corresponding arrival time results.

[0165] Preferably, in the multi-round expansion planning process, the proportion of remaining unfinished customer nodes is used to represent the degree of unfinished tasks in the current round, and it is used as an additional penalty item in the multi-round expansion task planning process to participate in the calculation of the comprehensive evaluation score in step S5.

[0166] The remaining percentage of uncompleted customer nodes is calculated as follows:

[0167]

[0168] in, Indicates the first The percentage of remaining uncompleted client nodes in an ant colony. Indicates the first The number of customer nodes that have not completed delivery in the ant colony solution.

[0169] The remaining uncompleted customer nodes are used as an additional penalty in the multi-round expansion task planning process. The overall evaluation score is then calculated as follows:

[0170]

[0171] in, Indicates the first The overall evaluation score of the ant colony solution. Indicates the first The overall objective value of the ant colony solution. This indicates the penalty weight for unfulfilled customer nodes.

[0172] Example 1

[0173] This invention provides a multi-drone delivery collaborative task planning method, applicable to delivery scenarios consisting of a delivery center and multiple customer nodes. Each customer node has corresponding spatial coordinates, task weight, and task priority information. Multiple drones depart from the delivery center to execute delivery tasks and return to the delivery center after completing their respective task sequences.

[0174] To address this scenario, this invention establishes a distance matrix between nodes and a heuristic information matrix, introduces a dynamic remaining range budget model to screen the feasibility of candidate delivery paths, generates multiple UAV candidate access behaviors based on a shared set of unvisited customer nodes, and completes the collaborative path construction of multiple UAVs by combining a single-step competition mechanism. On this basis, iterative optimization of the solution process is achieved through a comprehensive objective function, constraint penalty terms, ranking-based global pheromone updates, adaptive adjustment of volatile factors, and pheromone upper and lower bound pruning. When a single round of delivery does not cover all customer nodes, multiple rounds of extended planning are performed on the remaining customer nodes until the final delivery plan is output.

[0175] This task planning method is illustrated through a case study involving a distribution center, three drones, and 30 customer nodes. Information such as the location of the distribution center and customer nodes, and material requirements, is shown in Table 1.

[0176] Table 1. Location information, material requirements, and priority data for the central warehouse and each customer node.

[0177]

[0178] In this embodiment, node 0 is the distribution center with coordinates (5,5); nodes 1-30 are customer nodes. The coordinates, weight, and priority information of each node are detailed in Table 1. The weight represents the task weight of the customer node, and a higher priority value indicates a more urgent task.

[0179] The solution obtained by using the method of this invention for task planning is as follows: Figure 6 As shown in Table 2, the generated multi-round delivery routes are as follows:

[0180] Table 2 Results of Multi-Round Delivery Task Planning

[0181]

[0182] As can be seen from Table 2, the method of the present invention divides the delivery tasks of 30 customer nodes into 3 rounds for completion:

[0183] In the first round of delivery, drone 0 visited nodes 10→16→26→15 in sequence, drone 1 visited nodes 17→14→3→8 in sequence, and drone 2 visited nodes 27→13→1→25→19→24 in sequence.

[0184] In the second round of delivery, drone 0 visited nodes 11→9→18→5→28 in sequence, drone 1 visited nodes 4→23→12→2 in sequence, and drone 2 visited nodes 30→7→6→21→20→22 in sequence.

[0185] In the third round of delivery, drone 0 visited the remaining node 29 and completed all delivery tasks.

[0186] Figure 7 The convergence of the method of the present invention in the multi-round extended solution process is shown. It can be seen that the algorithm can converge effectively in each round, which verifies the effectiveness and stability of the method of the present invention in multi-UAV delivery task planning.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for collaborative task planning of multi-UAV delivery, characterized in that, Applicable to a distribution center, drones and This is a delivery scenario consisting of several customer nodes, where the node number of the delivery center is denoted as 0, and its coordinates are shown as follows: The node IDs of all client nodes constitute the client node ID set. For the first There are 1 client node, whose coordinates are: The task weight is recorded as Task priority is recorded as The maximum payload of each drone is The maximum equivalent range when unloaded is The maximum effective range when fully loaded is The method includes the following steps: Step S1: Obtain delivery task data and construct basic planning information: Obtain node information of the delivery center and all customer nodes, calculate the distance between customer nodes based on the coordinate information of each customer node, construct a distance matrix, construct a heuristic information matrix based on the distance matrix, and construct and initialize the pheromone matrix. Step S2: Screening feasible candidate customer node set based on dynamic remaining range budget: Establish dynamic remaining range budget model, combine the currently constructed path, candidate customer nodes and the returning distribution center to form a candidate delivery path record, and perform segment-by-segment remaining range budget update; if the final remaining range budget obtained after completing the budget update of all segments of the candidate delivery path is greater than or equal to zero and the load constraint is less than the maximum load, then the candidate delivery path is determined to be feasible, and the candidate customer node is added to the set of feasible candidate customer nodes of the UAV; Step S3: Construct ant colonies and generate candidate access targets based on transition probabilities: In each iteration, construct multiple ant colonies. Each ant colony is used to independently generate a set of multi-drone delivery paths. Based on the paths currently constructed by each drone in the ant colony, determine the set of visited customer nodes of the ant colony. Based on the shared set of unvisited customer nodes, determine the set of feasible candidate customer nodes for each drone. Calculate the transition probabilities based on the pheromone matrix and the heuristic information matrix to generate candidate access targets for each drone. Step S4: Update the multi-drone cooperative path based on the single-step competition mechanism to obtain the ant colony solution: For each drone that generates candidate access behavior, calculate its competition cost; select the drone with the minimum competition cost as the winner and execute the current step of delivery; after the winning drone completes the access edge selection, perform local pheromone update on the access edge. Step S5: Perform a comprehensive evaluation of the ant colony solutions and execute global pheromone updates, adaptive volatile factor adjustments, and pheromone pruning: Calculate the comprehensive objective value and constraint penalty for each ant colony solution to obtain the comprehensive evaluation score; select the ant colony with the best comprehensive evaluation score. Each optimal solution participates in the ranking-based global pheromone update; the volatile factor is adaptively adjusted, and the upper and lower bounds of the pheromone matrix are pruned. Step S6: Output the single-round delivery result and perform multi-round extended planning, and output the final delivery plan: Determine whether the termination condition is met; when the single-round delivery planning result does not cover all customer nodes, extract the remaining customer node set and reconstruct the subproblem, and repeat steps S1 to S5; repeat the extended planning process until all customer nodes have completed delivery, and output the final delivery plan.

2. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S1, the formula for calculating the elements of the heuristic information matrix is: in, Indicates the first The first customer node and the first Heuristic information between client nodes This represents the maximum distance in the distance matrix. Indicates the first Task priority weights for each client node This represents the sum of the priority weights of all client node tasks; when or At that time, corresponding heuristic information Set to 0.

3. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S2, the update formula for the remaining range budget is: in, This indicates the estimated remaining distance before the start of the current flight segment. This indicates the budget for the remaining distance after completing the current flight segment. Indicates the current flight distance. This indicates the remaining weight of the drone to be delivered at the start of the current flight segment. Indicates the maximum load capacity. Indicates the maximum equivalent range when unloaded. This indicates the maximum equivalent range when fully loaded.

4. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S3, let the first... The ant colony in the 1st The set of feasible candidate client nodes obtained after the drone traverses unvisited client nodes according to step S2 and completes the judgment of load constraints and remaining range budget is as follows: Under the constraint of sharing the set of unvisited customer nodes, the set of feasible candidate customer nodes for the UAV to participate in the transfer probability calculation is defined as follows: in, Indicates the first The set of visited client nodes of an ant colony; the transition probability satisfies: in, For the first The current set of feasible candidate customer nodes for drones Candidate customer nodes in the process, Represents client node To candidate customer node pheromone intensity, Represents client node To candidate customer node Heuristic information, Represents client node To candidate customer node pheromone intensity, Represents client node To candidate customer node Heuristic information, Represents pheromone factor, Represents the heuristic factor.

5. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S4, the formula for calculating the competition cost is: in, Indicates the first The ant colony in the 1st The competitive cost of a drone in the current path construction step; Indicates the first The drone's current location is compared to the candidate access targets generated for the drone in step S3. distance, Indicates the candidate access target Distance to the distribution center Indicates candidate delivery routes The remaining range budget is obtained after updating the remaining range budget for each segment. Indicates the first The ant colony in the 1st The total load on the corresponding path after a drone adds candidate customer nodes to the current path. and The competition cost weighting coefficient, To prevent tiny positive numbers with a denominator of zero.

6. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S4, the local pheromone update formula is: in, Indicates time Side pheromone concentration, Indicates time side The concentration of pheromones on the surface Indicates the local pheromone volatile factor. This indicates the initial pheromone concentration.

7. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S5, the formula for calculating the comprehensive target value is: in, Indicates the first The overall objective value of the ant colony solution. This represents the ant colony index. Indicates the drone index; This represents the length of the complete mission execution path of the u-th drone in the k-th ant colony. Indicates the baseline path length. This represents the priority-weighted arrival cost of the u-th drone in the k-th ant colony. Indicates the number of client nodes. This represents the sum of the task priority weights of all client nodes. and These represent the weighting coefficients for the path length item and the priority-weighted timeliness item, respectively. The formula for calculating the comprehensive evaluation score is as follows: in, Indicates the first The overall evaluation score of the ant colony solution. Indicates the first The total range penalty for an ant colony to resolve Indicates the first The total load penalty for solving an ant colony. and These represent the range penalty weight and the load penalty weight, respectively.

8. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S5, the ant colony solutions are sorted from smallest to largest based on their comprehensive evaluation scores, and the top scores are selected. The optimal solutions are selected as candidate solutions to participate in the global pheromone update, and their ranking weights are calculated: in, Indicates the first The ranking weight of each optimal solution; Let the first The set of path edges contained in each optimal solution is The preferred solution lies on the edge. pheromone increment Defined as: in, Indicates the initial pheromone concentration. Indicates the first The comprehensive evaluation score of each optimal solution; side The pheromones on the surface are updated globally in a ranked manner as follows: in, Indicates time Side pheromone concentration, Indicates time side The concentration of pheromones on the surface Indicates the first Global pheromone evaporation factor at the next iteration.

9. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In step S5, the adaptive adjustment formula for the volatile factor is: in, Indicates the first The local pheromone evaporation factor used in this iteration Indicates the first The global pheromone evaporation factor used in this iteration Indicates the first The local pheromone evaporation factor used in this iteration Indicates the first The global pheromone evaporation factor used in this iteration and These represent the lower bounds of the local volatile factor and the global volatile factor, respectively. and These represent the attenuation coefficients of the local volatile factor and the global volatile factor, respectively. After completing the global pheromone update, the pheromone matrix is ​​pruned using upper and lower bounds. The pruning formula is as follows: in, This represents the edges after ranking-based global pheromone updates and upper and lower bound clipping. pheromone concentration, and These represent the lower and upper bounds of the permissible pheromone concentration, respectively.

10. The multi-UAV delivery collaborative task planning method according to claim 1, characterized in that, In the multi-round extended task planning process based on the sub-problem of reconstructing remaining unfinished customer nodes, the proportion of remaining unfinished customer nodes is used to represent the degree of unfinished tasks in the current round, and this proportion is used as an additional penalty item in the comprehensive evaluation score calculation during the multi-round extended task planning process. in, Indicates the first The percentage of remaining uncompleted client nodes in an ant colony. Indicates the first The number of customer nodes that have not completed delivery in the ant colony solution. Indicates the total number of customer nodes. Indicates the first The overall evaluation score of the ant colony solution. Indicates the first The overall objective value of the ant colony solution. Indicates the first The total range penalty for an ant colony to resolve Indicates the first The total load penalty for solving an ant colony. and These represent the range penalty weight and the load penalty weight, respectively. This indicates the penalty weight for unfulfilled customer nodes.

Images