Heterogeneous unmanned aerial vehicle task allocation and reallocation method based on alliance formation game

Abstract

The invention discloses a heterogeneous unmanned aerial vehicle task allocation and reallocation method based on an alliance formation game. The method comprises the following steps: respectively clustering unmanned aerial vehicles and tasks to obtain an unmanned aerial vehicle class cluster and a task class cluster; for each unmanned aerial vehicle class cluster and the corresponding task class cluster, the method comprises the following steps: forming a game model based on an alliance, and constructing a revenue function, including a resource contribution function, a path cost function and a threat cost function, of an unmanned aerial vehicle under each task alliance; based on a Nash equilibrium theory in the game theory, modeling a task allocation problem as a task alliance division problem of a maximized potential function according to a revenue function; solving a task alliance division problem of a maximized potential function based on processing of two stages of a maximum weighted matching stage and an unmanned aerial vehicle transfer alliance judgment stage, and realizing task allocation of the unmanned aerial vehicle; and when a certain unmanned aerial vehicle fails, tasks of the failed unmanned aerial vehicle are redistributed. According to the method, the task allocation efficiency, the system robustness and the fault-tolerant capability are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned system collaborative control and resource optimization, and specifically relates to a heterogeneous unmanned aerial vehicle task allocation and reallocation method based on alliance formation game. Background Art

[0002] With the development of drone technology, heterogeneous drone swarms are increasingly being used in various fields. For example, in some scenarios, multiple types of drones collaborate to perform reconnaissance, jamming, and strike missions. In other scenarios, tasks such as logistics distribution and emergency rescue require efficient task allocation strategies.

[0003] Currently, UAV task allocation primarily relies on centralized optimization algorithms, heuristic algorithms, intelligent algorithms, and game theory methods. Optimization theory methods seek optimal solutions through mathematical modeling such as linear programming and integer programming, but these methods are computationally complex and struggle to meet real-time requirements. Heuristic algorithms, such as genetic algorithms and ant colony algorithms, are computationally efficient but prone to local optima and have limited ability to model heterogeneity. Intelligent algorithms, such as deep learning and reinforcement learning, offer strong dynamic adaptability but require long training cycles and consume significant resources. Game theory methods, such as coalition formation games, coordinate multi-agent interactions, but are insufficiently adaptable to heterogeneity and lack dynamic reallocation mechanisms. Furthermore, all existing heterogeneous UAV task allocation methods lack dynamic adaptability and lack rapid reallocation mechanisms in the event of UAV failure or environmental changes, increasing the risk of mission failure. Summary of the Invention

[0004] In order to solve the above problems existing in the prior art, the present invention provides a method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0005] An embodiment of the present invention provides a method for allocating and reallocating tasks of heterogeneous drones based on an alliance formation game, the method comprising:

[0006] Based on the drone status information, all drones are clustered to obtain a number of drone clusters, and based on the task status information, all tasks are clustered to obtain a number of task clusters, where the drone clusters correspond one to one to the task clusters;

[0007] For each drone cluster and corresponding task cluster, the execution steps include: initializing the task alliance and modeling a coalition formation game model including a drone set, a task set, a payoff function for evaluating all drones in the task alliance, and evaluating the utility of individual drones in each task alliance; based on the coalition formation game model, constructing a payoff function for each drone in the task alliance, including a resource contribution function, a path cost function, and a threat cost function; based on the Nash equilibrium theory in game theory, modeling the task allocation problem as a task alliance partitioning problem that maximizes a potential function according to the payoff function; solving the task alliance partitioning problem that maximizes the potential function based on a two-stage process of a maximum weighted matching stage and a drone transfer alliance judgment stage, and obtaining a Nash equilibrium coalition structure to achieve drone task allocation;

[0008] When a drone fails, the Nash equilibrium alliance structure of the failed drone is determined, and the failed drone is removed from the determined Nash equilibrium alliance structure, and the task corresponding to the failed drone is used as the task to be assigned; in the Nash equilibrium alliance structure after the failed drone is removed, the insertion positions of multiple tasks to be assigned are determined to form several candidate task alliances; based on the two-stage processing of the maximum weighted matching stage and the drone transfer alliance judgment stage, the task alliance partitioning problem of maximizing the potential function corresponding to several candidate task alliances is solved to select the optimal candidate task alliance from all candidate task alliances, and the optimal candidate task alliance is added to the Nash equilibrium alliance structure to realize the task reallocation of the failed drone.

[0009] Beneficial effects of the present invention:

[0010] The present invention proposes a heterogeneous UAV task allocation and reallocation method based on alliance formation game. In order to improve the efficiency of task allocation, the UAVs and tasks are first intelligently grouped, and the cluster centers are dynamically adjusted, thereby improving the matching accuracy of heterogeneous UAVs and tasks, reducing resource waste, and improving the overall task execution efficiency. By modeling an alliance formation game model, a distributed task allocation architecture is realized. This distributed task allocation architecture reduces communication and computing overhead and is more suitable for large-scale UAV cluster task scheduling than traditional centralized algorithms. It enables UAVs to collaboratively perform tasks, reduces task conflicts, and can adaptively adjust task allocation in a dynamic environment, thereby improving overall execution efficiency. By designing a benefit function, the contribution, path cost and risk factors of UAVs are comprehensively considered to make task allocation more reasonable. The two-stage processing strategy of the maximum weighted matching stage and the UAV transfer alliance judgment stage is combined to improve computing efficiency and ensure that task allocation can be completed within a reasonable time. The task reallocation mechanism is designed to ensure that the tasks of faulty UAVs can be quickly transferred to healthy UAVs to avoid task interruption. In general, the present invention combines the complete process of clustering preprocessing, game theory decision-making, distributed task allocation and reallocation, ensuring that the task allocation scheme adapts to dynamic environmental changes and is suitable for multiple application scenarios such as logistics and emergency rescue. It enhances the versatility and practicality of the task allocation scheme and has wider application value.

[0011] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a flow chart of a method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0014] This invention aims to solve the resource heterogeneity, dynamic adaptability and real-time performance issues in task allocation for heterogeneous UAV clusters, and improve task allocation efficiency, system robustness and fault tolerance. Figure 1 The embodiment of the present invention provides a method for allocating and reallocating tasks of heterogeneous UAVs based on alliance formation game, which specifically includes the following steps:

[0015] S10. Based on the drone status information, cluster all drones to obtain several drone clusters, and based on the task status information, cluster all tasks to obtain several task clusters, with drone clusters and task clusters corresponding one to one.

[0016] In the embodiment of the present invention, there are n drones in the drone cluster, and the drone set is represented by U = {u1, u2, ... u n These drones are heterogeneous in terms of type and resources, meaning they can provide different performance when performing the same mission. This is quantified by the type and amount of resources they carry. The time required for drones to change flight direction and speed is negligible. Therefore, the state information defined for each drone includes drone resources, drone position, drone value, and drone speed, and is defined as:

[0017]

[0018] in, Indicates drone u i Status information, i ranges from 1 to n, Indicates drone u i resources, Indicates drone u i location, Indicates drone u i The value of Indicates drone u i The flight speed of the UAV is limited, so the flight speed constraint is introduced. Indicates drone u i Maximum flight speed; Indicates drone u i The lth type of resource, l represents the total number of resource types; because the value of drones themselves is different, the value information is introduced To improve the applicability of the model.

[0019] In this embodiment of the present invention, there are m tasks in total, and the task set is represented as: T = {t1, t2, ..., t m Each mission contains different types of subtasks. If the UAV arrives at the mission later than a certain time, the mission fails. The status information defined for each mission includes the number of UAV resources required for the mission, the mission location, the mission hard time window information, and the mission threat level, which is defined as:

[0020]

[0021] in, Represents task t j Status information, j ranges from 1 to m, Represents task t j The number of drone resources required, Represents task t j location, Represents task tj Hard time window information, Represents task t j Threat level; Represents task t j The number of resources of type K required, K ranges from 1 to 1, and the specific value of K depends on the task t j Demand for drone resources; Represents the hard time window information of the task, where Indicates that the UAV arrives at task t j The earliest time, Indicates that the UAV arrives at task t j The latest time; the mission itself may have a certain degree of damage to the drone, so the threat level is introduced Indicates that task t j Threat factor to drones.

[0022] In the embodiment of the present invention, the intersection of alliances corresponding to different tasks is an empty set, and the drone alliance can execute at most one task.

[0023] Furthermore, to improve task allocation efficiency, a dynamic clustering algorithm is first used to cluster and group drones and tasks. This embodiment of the present invention uses a K-medoids algorithm to cluster drones and tasks, checking whether the clustering results meet the set conditions. If not, the cluster centers are reinitialized and an iteration is performed, thereby achieving reasonable clustering of drones and tasks based on heterogeneous resources.

[0024] The embodiment of the present invention clusters all drones based on drone resources and drone positions to obtain several drone clusters, including: dividing all drones into several drone clusters; for each drone cluster, executing steps including: calculating the first proportion of each drone in the drone cluster under each type of resource based on drone resources; calculating the resource entropy value of the corresponding drone based on all first proportions, selecting the drone position corresponding to the largest resource entropy value from all resource entropy values as the drone cluster center, calculating the first distance error between each drone in the drone cluster and the drone cluster center, judging whether the sum of all first distance errors meets a first preset threshold, and if so, calculating the resource balance of the drone cluster, judging whether the resource balance of the drone cluster is less than or equal to a second preset threshold, and if so, indicating that each type of resource in the drone cluster meets the balance requirement, and if not, returning to the step of dividing all drones into several drone clusters until each type of resource in all drone clusters meets the balance requirement. More specifically:

[0025] Step 11, input parameter initialization: input the drone resources of each drone in the drone set and location (UAV u i As an example), let the number of drone clusters be k1, and the k1 drone clusters are recorded as C u ={c1,c2,...,c k1}, each drone cluster includes multiple drones;

[0026] Step 12, cluster center selection: For each drone in each drone cluster, calculate the first proportion of each type of resource of the drone in the total resources. The calculation formula of the first proportion is expressed as:

[0027]

[0028] in, Indicates drone u i The first proportion of the lth type of resources in the total resources, Indicates drone u i The available quantity or occupancy of the first type of resources, Indicates drone u i The available quantity or occupancy of all resource types.

[0029] Next, the resource entropy value of the corresponding drone is calculated based on all the first proportions. The formula is expressed as:

[0030]

[0031] Among them, Entropy(u i ) indicates drone u i After calculating the resource entropy values corresponding to all drones, the location of the drone with the largest resource entropy value is selected as the drone cluster center, denoted as o u .

[0032] Step 13, main iterative optimization: In order to evaluate the effect of the clustering results, the first distance error between each drone in the drone cluster and the drone cluster center is calculated. The formula is expressed as:

[0033]

[0034] Among them, E u represents the first distance error, represents the k1th drone cluster, Indicates that drone u belongs to the drone cluster Inside, p u Denotes the location of drone u. Determine whether the sum of all first distance errors in the drone cluster meets the first preset threshold. If so, calculate the resource balance of the drone cluster. Indicates the available quantity or occupancy of the lth type of resource of UAV u, Express The result after normalization processing is used to determine whether the resource balance of the drone cluster is less than or equal to the second preset threshold, that is, to determine the balance of each type of resource in the drone cluster. If so, it indicates that each type of resource in the drone cluster meets the balance requirements. If not, there may be a problem of uneven resource distribution, and the drone cluster structure needs to be adjusted. Then return to the step of dividing all drones into several drone clusters, reselect k1 drone clusters, perform new clustering, and continue to check resource balance until each type of resource in all drone clusters meets the balance requirements. This process is repeated to ensure that the final drone cluster is more balanced in resource distribution.

[0035] The embodiment of the present invention clusters all tasks based on the number of drone resources required for the tasks and the task locations to obtain several task clusters, including: dividing all tasks into several task clusters; the number of task clusters is equal to the number of drone clusters; for each task cluster, executing steps including: calculating the second proportion of the number of each type of drone resources required for each task in the task cluster based on the number of drone resources required for the task to the total number of all types of drone resources required for the task; calculating the resource demand entropy value of the corresponding task based on all second proportions, selecting the task location corresponding to the largest resource demand entropy value from all resource demand entropy values as the task cluster center, calculating the second distance error between each task in the task cluster and the task cluster center, judging whether the sum of all second distance errors meets the third preset threshold, and if so, calculating the resource balance of the task cluster, judging whether the resource balance of the task cluster is less than or equal to the fourth preset threshold, and if so, indicating that each type of resource in the task cluster meets the balance requirement, and if not, returning to the step of dividing all tasks into several task clusters until each type of resource in all task clusters meets the balance requirement. More specifically:

[0036] Step 21, input parameter initialization: input the number of drone resources required for each task in the task set and location (Task t j For example), the number of task clusters is set to k1, that is, the number of task clusters and drone clusters is consistent, then k1 task clusters are recorded as C t ={cc1,cc2,...,cc k1}, each task cluster includes multiple tasks;

[0037] Step 22, cluster center selection: For each task in each task cluster, calculate the second ratio of the number of drone resources required for each task in the task cluster to the total number of all types of drone resources required for the task. The second ratio calculation formula is expressed as:

[0038]

[0039] in, Represents task t j The number of required UAV resources of type k is the second largest proportion of the total number of all types of UAV resources required for the mission. Represents task t j The number of k-th type of drone resources required, Represents task t j The number of drone resources of all types required.

[0040] Next, the resource requirement entropy value of the corresponding task is calculated based on all the second proportions. The formula is expressed as:

[0041]

[0042] Among them, Entropy(t j ) represents task t j After calculating the resource demand entropy values corresponding to all tasks, the task location with the largest resource demand entropy value is selected as the task cluster center, denoted as o t .

[0043] Step 23, main iterative optimization: In order to evaluate the effect of the clustering results, the first distance error between each drone in the drone cluster and the drone cluster center is calculated. The formula is expressed as:

[0044]

[0045] Among them, E t represents the second distance error, represents the k1th task cluster, Indicates that task t belongs to the task cluster Inside, p t Denotes the location of task t. Determine whether the sum of all second distance errors in the task cluster meets the third preset threshold. If so, calculate the resource balance of the task cluster. represents the number of k-th type resources required by task t, Express The result after normalization processing is used to determine whether the resource balance of the task cluster is less than or equal to the fourth preset threshold, that is, to determine the balance of resources required for each task in the task cluster. If so, it indicates that the resources required for each task in the task cluster meet the balance requirements. If not, there may be a problem of uneven resource distribution, and the task cluster structure needs to be adjusted. Then return to the step of dividing all tasks into several task clusters, reselect k1 task clusters, perform new clustering, and continue to check the balance of required resources until the resources required for each task in all task clusters meet the balance requirements. This process is repeated to ensure that the final task cluster is more balanced in terms of resource demand distribution.

[0046] The resulting clustering of drone clusters and mission clusters creates a one-to-one correspondence, forming a complete partition. When the total resources of each drone cluster meet mission requirements and the resources for individual drones and individual missions are relatively balanced, the drone resources within the partition after resource-based clustering can also meet mission requirements.

[0047] S20. For each drone cluster and the corresponding task cluster, the execution steps include: initializing the task alliance, and modeling an alliance formation game model including a drone set, a task set, a benefit function for evaluating all drones under the task alliance, and an evaluation of the utility of a single drone under each task alliance; based on the alliance formation game model, constructing a benefit function of the drone under each task alliance including a resource contribution function, a path cost function, and a threat cost function; based on the Nash equilibrium theory in game theory, modeling the task allocation problem as a task alliance division problem that maximizes the potential function according to the benefit function; based on the two-stage processing of the maximum weighted matching stage and the drone transfer alliance judgment stage, solving the task alliance division problem that maximizes the potential function, and obtaining a Nash equilibrium alliance structure to realize the task allocation of drones.

[0048] The embodiment of the present invention models the UAV swarm task allocation problem as a coalition formation game model. The coalition formation game model is defined as: G = (U, E, ε, R), where the UAV set U = {u1, u2, ..., u n}, task set T = {t1, t2, ..., t m}, where the strategy set E is the space of optional mission strategies for all drones, which is equivalent to the mission set T, ε is the utility function used to evaluate drone benefits, and R is used to evaluate the utility of each alliance. Therefore, when each drone in the drone set makes a strategy, (m+1) mission alliances will be formed. The goal of the game is for the drone set to select the appropriate strategy to obtain a stable alliance structure. It can be seen that the stable alliance structure includes the UAV alliance corresponding to m tasks. Represents task t mThe corresponding drone alliance is assigned to perform the task t j The collection of all drones. Among them, Represents a drone alliance that has not been assigned a task.

[0049] The profit function constructed in the embodiment of the present invention is expressed as follows:

[0050]

[0051] in, Indicates drone u i In the task j The profit under i ,t j ) indicates drone u i In the task j The resource contribution function under COST(u i ,t j ) indicates drone u i In the task j The path cost function under RISK(u i ,t j ) indicates drone u i In the task j The threat cost function under , α, β, and γ are all constant weight values, which respectively determine the resource overlap and path cost and the proportion of the threat cost of executing the task in the benefit;

[0052] Resource contribution function VAL(u i ,t j ), the formula is:

[0053]

[0054] Among them, η and λ represent constant weights respectively. η is used to weigh the importance of heterogeneous resources in the task, and λ is used to weigh the proportion of resources in the benefits. RN represents the number of drones u i Join the task j The number of resources available in the corresponding task alliance, Indicates drone u i Join the task j The number of unused resources in the corresponding task alliance, i.e., redundant resources;

[0055] Path cost function COST(u i ,t j ), the formula is:

[0056]

[0057] Among them, d(u i ,tj ) indicates drone u i With the task t j The Euclidean distance between them can be obtained by the drone position and task location To calculate the Euclidean distance between them, Represents task t j The size of the task environment area, μ represents a constant less than 0;

[0058] Threat cost function RISK(u i ,t j ), the formula is:

[0059]

[0060] in, Indicates drone u i The value of Represents task t j The threat level of RISK(u i ,t j ) is used to express the threat cost assessment of mission factors on drones. The mission will cause damage to drones. By taking into account the value of heterogeneous drones, the overall loss of drones can be reduced.

[0061] Based on the characteristics of different tasks, the priority of tasks can be balanced through the profit function. For example, fresh food tasks emphasize temperature control matching, medical tasks prioritize timeliness, and ordinary package tasks allow a certain time error. This allows for comprehensive consideration of factors such as load, temperature control, and timeliness when allocating tasks.

[0062] Task t in a stable alliance structure j Corresponding drone alliance The coalition utility is shown as: Then the total revenue of task allocation of all drone alliances in a stable alliance structure is:

[0063]

[0064] The utility of a single drone is evaluated as:

[0065]

[0066] in Indicates drone u i Select a strategy Joined the mission alliance, Indicates that the strategy The remaining alliance utility after being removed from the original task alliance.

[0067] The goal of the game of this invention is to obtain the final stable Nash equilibrium solution CS * Nash equilibrium is a concept of solution in game theory. Nash equilibrium refers to a strategy combination that satisfies the following properties: under the premise that the strategies of other players remain unchanged, any player will not increase his or her own benefits if he or she unilaterally changes his or her own strategy in this strategy combination. In this combination, the strategy of each player is to maximize his or her expected benefits. In the embodiment of the present invention, Nash equilibrium CS * Specifically defined as:

[0068]

[0069] in, represents the Nash equilibrium strategy of drones in a stable alliance structure, Indicates drone u i The utility, Indicates drone u i The strategy set of other drones in a stable alliance structure, Indicates except Drones constantly switch between alliances to improve their own utility. When no drone can improve its own benefits by adjusting its alliance independently, a stable alliance structure is achieved. The goal of the alliance formation game is to achieve a stable alliance structure, which is consistent with the goal of task allocation. Therefore, the solution obtained through this game is the solution to the task allocation problem.

[0070] In a potential game, each participant can adjust their strategy to maximize their own benefits or utility. If there exists a potential function P that satisfies:

[0071]

[0072] Among them, E u represents the strategy set of all drones in the alliance structure, Indicates drone u i The strategy set of other drones in the alliance structure, It's a drone i The strategy of choice, Indicates drone u i Except for the selected strategy If any strategy other than the above is able to reflect the change in utility after the participants change their strategies, then the game is called a potential game. Because potential games must have a Nash equilibrium point, the Nash equilibrium of the game can be achieved by minimizing or maximizing the potential function. Under the finite increment characteristic, the potential game can converge to the Nash equilibrium state in a finite time. If the utility function change of a certain drone joining or exiting a certain alliance structure alone is reflected in the potential function when the strategies of other drones remain unchanged, the potential game model can be used to obtain the optimal solution. Therefore, the embodiment of the present invention constructs the established alliance formation game model as a potential game, and the Nash equilibrium solution of the alliance formation game model exists. That is: When the drone i The strategy of becomes at this time:

[0073]

[0074] in, Does not include strategy and The change of UAV strategy is represented in the alliance structure as follows: So we can get:

[0075]

[0076] In other words, there exists a potential function that can equally reflect the changes in the strategies of individual drones, and the alliance formation game model of the present invention can be constructed as a potential game. is the potential function. Moreover, due to the nature of potential games, the Nash equilibrium solution of the game model must exist, and a reasonable alliance formation game algorithm can be designed to achieve the final task allocation.

[0077] The objective function of the UAV during task assignment can be specifically expressed as:

[0078]

[0079] At the same time, when implementing heterogeneous drone allocation and reallocation in the embodiment of the present invention, the following constraints must be met:

[0080] UAV flight speed constraints:

[0081] Task timeliness constraints: Indicates the task deadline, mint(u i ,t j ) indicates drone u i Arrival mission t j The shortest time, that is, the drone u iNeed to arrive at task t before the deadline j .

[0082] Constraints on the UAV mission execution mode: The intersection of UAV alliances corresponding to different tasks is an empty set, and a UAV can execute at most one task.

[0083] Based on the above analysis, the present invention can model the task allocation problem as a coalition partitioning problem that maximizes a potential function. This demonstrates that the present invention models task allocation as a coalition formation game, enabling heterogeneous drones to collaboratively complete tasks. The designed game utility function comprehensively considers drone contributions, task benefits, path costs, and risk factors to achieve a stable coalition structure (Nash equilibrium). Furthermore, the task allocation game is proven to be a potential game, ensuring the convergence and stability of the algorithm.

[0084] Based on this analysis, the embodiment of the present invention performs task allocation on each cluster partition obtained after clustering drones and tasks, and obtains a solution for local allocation within each cluster partition. When all cluster partitions are allocated, global task allocation is completed. The embodiment of the present invention proposes: first, a solution is obtained based on the maximum weighted matching principle, and then drone transfer alliance rules are set to establish a stable alliance structure until the drone alliance resources corresponding to all tasks meet the demand. Specifically:

[0085] The embodiment of the present invention solves the task alliance partitioning problem of maximizing the potential function based on the two-stage processing of the maximum weighted matching stage and the drone transfer alliance judgment stage, and obtains the Nash equilibrium alliance structure, including:

[0086] In the maximum weighted matching phase, a bipartite graph is constructed for each task alliance, and maximum weighted matching is performed on the bipartite graph to obtain the corresponding initial allocation result. The construction of the bipartite graph for each task alliance includes: the task set and drone set in each task alliance are used as the first set and the second set in the bipartite graph, and the maximum value of the benefit of each drone after joining the corresponding task alliance is used as the edge weight in the bipartite graph.

[0087] In the drone transfer alliance judgment stage, the drone transfer alliance rules are designed. According to the initial allocation results and the drone transfer alliance rules, the drone transfer alliance is made to reach a stable state, and all drone transfer alliances are completed to obtain a Nash equilibrium alliance structure; the design of the drone transfer alliance rules includes adding the drones allocated in the initial allocation results to other task alliances for utility comparison. When the utility of the drone joining other task alliances is greater than the utility of the drone in the original task alliance, the drone withdraws from the original task alliance and waits for reallocation. Otherwise, the drone has reached a stable state.

[0088] The process of constructing a bipartite graph for each task alliance in the embodiment of the present invention specifically includes:

[0089] Mark all tasks in each task alliance as unassigned, all drones are not assigned tasks, and the task alliance corresponding to each task is an empty set; calculate the benefits of each drone in the task alliance when performing each task individually to obtain a benefit matrix; convert the benefit matrix into a bipartite graph, and perform maximum weighted matching on the bipartite graph to obtain the corresponding initial allocation result; the conversion and matching process of the bipartite graph includes: each drone in the drone set is regarded as a vertex in the first set of the bipartite graph, and the value of each vertex in the first set is 0; each task in the task set is regarded as a vertex in the second set of the bipartite graph, and the value of each vertex in the second set is the maximum benefit of each drone after joining the task alliance individually, and use the idea of the Hungarian algorithm to perform maximum weighted matching on the bipartite graph to obtain the initial allocation result. More specifically:

[0090] Step 31, initialize the allocation result: m tasks in the task alliance are marked as unassigned, and the corresponding drone alliance is an empty set; all drones are not assigned tasks;

[0091] Step 32: When the UAV alliance corresponding to each task is an empty set, calculate the benefits of each UAV in the task alliance performing each task individually according to formula (9) to obtain the M×N dimensional benefit matrix V. In order to meet the timeliness constraints of the task, when calculating the benefit matrix of the UAV joining the task, it is necessary to determine whether the constraint conditions can be met based on the distance between the UAV and the task;

[0092] Step 33: Convert the payoff matrix into a bipartite graph. The first set X on the left represents the tasks, and each task node corresponds to a top label x[i]. This top label x[i] is initially set to the maximum payoff of each drone after individually joining the task alliance, that is, the maximum payoff among all drones in the task selection. The drone vertex is preset to y[i] = 0. Begin matching. The matching principle is that the sum of the vertex values must equal the weight ω[i][j] of the edge connecting to it. The weight ω[i][j] corresponds to the payoff matrix V[i][j]. Utilizing the Hungarian algorithm, the top labels are continuously modified and deep exploration is performed to find feasible augmenting paths, ultimately achieving maximum weighted matching and obtaining the initial allocation result.

[0093] The process of designing the drone transfer alliance rules in the embodiment of the present invention specifically includes: calculating the first utility of the drone in the original task alliance; calculating the second utility of the drone joining other task alliances; comparing whether the second utility is greater than the first utility. If so, the drone exits the original task alliance and waits for reallocation; if not, the drone has reached a stable state and remains in the original task alliance; calculating the benefits of the remaining drones in the task alliance when they perform each remaining task individually to obtain a new benefit matrix, returning to the step of converting the benefit matrix into a bipartite graph, and performing maximum weighted matching on the bipartite graph to obtain the corresponding initial allocation result, until all drones reach a stable state. More specifically:

[0094] Step 41: Calculate the first utility of the drone in the original mission alliance: in Indicates drone u i Select a strategy When joining the original task alliance, Is the drone u i The coalition utility of the remaining drones in the original mission alliance after being removed from the original mission alliance;

[0095] Step 42: Calculate the second utility of the drone joining other mission alliances: in Indicates drone u i Join a mission alliance Then the alliance utility of the task alliance is calculated. Indicates drone u i Not joined the mission alliance The alliance utility before, that is, the original benefits of the system;

[0096] Step 43: Compare the size, if At this point, the drone exits the original mission alliance and awaits reallocation. Otherwise, the drone is considered stable and its original alliance remains unchanged. The drones that have reached a stable state are assigned to their mission alliances. The remaining set of unassigned drones is updated with the type and number of resources remaining after matching each task. A new payoff matrix is calculated, and the maximum weighted matching process described above is repeated based on the new payoff matrix. The newly matched drones are also subjected to utility checks according to the aforementioned drone alliance transfer rules, and the process is repeated in sequence. The convergence result is that all drones are stable, meaning that after the utility check, they will not increase their payoff by changing their mission alliances.

[0097] The embodiment of the present invention adopts game theory modeling and task alliance mechanism to realize a distributed task allocation strategy. Specifically, a local maximum weighted matching algorithm is used for preliminary task allocation to improve computational efficiency. Then, a UAV transfer alliance judgment mechanism is designed to allow UAVs to dynamically adjust their task alliances, thereby improving task execution efficiency. By iteratively optimizing the allocation results, all UAVs reach a stable state and ultimately obtain a global optimal solution, making task allocation more stable. Even in the event of UAV damage or environmental changes, it can still respond to changes such as UAV failures and new tasks in real time, maintaining the stability of task allocation and improving scheduling flexibility.

[0098] S30. When a certain UAV fails, determine the Nash equilibrium alliance structure in which the failed UAV is located, remove the failed UAV from the determined Nash equilibrium alliance structure, and use the task corresponding to the failed UAV as the task to be assigned; in the Nash equilibrium alliance structure after the failed UAV is removed, determine the insertion positions of multiple tasks to be assigned to form several candidate task alliances; based on the two-stage processing of the maximum weighted matching stage and the UAV transfer alliance judgment stage, solve the task alliance partitioning problem of maximizing the potential function corresponding to several candidate task alliances, so as to select the optimal candidate task alliance from all candidate task alliances, and add the optimal candidate task alliance to the Nash equilibrium alliance structure to realize the task redistribution of the failed UAV.

[0099] The embodiment of the present invention has completed the allocation of all tasks in a fault-free situation, and then each drone performs the task according to its strategy. A drone may suffer a permanent failure during the execution of a task, so that the faulty drone can no longer perform the remaining tasks. When a drone suffers a permanent failure, the terminated tasks need to be reallocated. The key to reallocation is how to transfer the tasks of the faulty drone to a healthy (normal) drone without affecting other tasks. When a failure occurs, the terminated tasks will be removed and then reinserted into the strategy of the healthy drone based on the priority of the remaining tasks. After removing the faulty task, the execution time of other tasks may need to be adjusted because time constraints will affect their utility. Then, the location of reinsertion is determined by splitting the executable time window of the task, so that the appropriate time point can be found for the healthy drone to perform the reallocated tasks without affecting other tasks. Specifically:

[0100] The embodiment of the present invention determines the insertion positions of multiple tasks to be assigned and forms several candidate task alliances, including: recalculating the earliest executable time of the task to be assigned based on the time when the faulty drone fails; dividing the task execution timeline based on the earliest executable time to obtain a predecessor task window, a successor task window, and an independent time window; determining multiple insertion positions that meet constraint conditions within the predecessor task window, the successor task window, or the independent time window, and inserting the task to be assigned into the insertion positions to obtain several candidate task alliances. More specifically:

[0101] The embodiment of the present invention first performs fault detection and task removal: identifying unfinished tasks caused by drone failures, removing the faulty drones and their tasks from the Nash equilibrium alliance structure, and generating a set of tasks to be reallocated.

[0102] When a permanent failure occurs in a UAV, the system will respond quickly and terminate the mission it is performing immediately. Let the time when the failure occurs be t f , the set of faulty drones is: F = {u f |u f Permanent failure occurs}, the set of unfinished tasks being executed by all failed drones is: in, It means to traverse all the drones belonging to the fault drone set F and put each fault drone u f The corresponding task sets are merged. Indicates a faulty drone u f Mission t has been reached j , that is, start executing the task, and the fault moment Indicates that it is earlier than the latest start time of the task This indicates that the task has not timed out and needs to be reassigned.

[0103] Furthermore, the terminated task is removed from the strategy of the faulty drone and marked as a task to be reallocated. f and the tasks they participate in j , task t j The remaining resource requirements are updated to

[0104] Then, the execution time window of subsequent tasks is adjusted, and the earliest executable time of the affected tasks is recalculated according to the time when the fault occurs.

[0105] Assume that the last task executed before the failure is t pre The end time is The first task after the failure t suc The earliest start time is updated to: in, Indicates drone u k From the fault time t f The first task after the failure t suc Flight time, Indicates the execution of task t suc The Drone Alliance.

[0106] In order to find a suitable location to reinsert the task to be reallocated, the task execution timeline is divided into the predecessor task window, the successor task window, and the independent time window.

[0107] The predecessor task window is the task interval before the fault, which is defined as:

[0108]

[0109] The subsequent task window is the cancellation interval of the faulty task, which is defined as:

[0110]

[0111] The independent time window is also the conflict-free interval, which is defined as:

[0112] in,

[0113] All windows must meet the following requirements:

[0114] Finally, the insertion positions of multiple tasks to be assigned are determined based on the three windows defined above, and the time t for each insertion position is insert ∈W pre ∪W suc ∪W ind The following constraints must be met:

[0115] Synchronicity constraint: The mission can only be started if multiple drones arrive at the same time;

[0116] Time constraint: The task needs to be completed before the deadline, that is,

[0117] Flight time constraints:

[0118] Through the above operations, multiple insertion positions that meet the constraint conditions are determined in the predecessor task window or the successor task window or the independent time window, and the tasks to be assigned are inserted into the insertion positions to obtain several candidate task alliances.

[0119] Furthermore, based on the obtained candidate task alliances, the task alliance partitioning problem corresponding to the candidate task alliances is solved based on the two-stage processing of the maximum weighted matching stage and the drone transfer alliance judgment stage, so as to select the optimal candidate task alliance from all candidate task alliances and add the optimal candidate task alliance to the Nash equilibrium alliance structure to achieve task redistribution of the faulty drone. Specifically:

[0120] In the embodiment of the present invention, the set of tasks to be reallocated is recorded as: T r ={t f ∈T|t f The mission alliance is affected by the fault and cannot be completed}, the healthy drones (drones that have eliminated the fault) are grouped as U h =U|F, where is the set of faulty drones. Construct a bipartite graph with the left task node set T r , the drone node set on the right is U h , the edge weight is the benefit of the drone joining the candidate task alliance. The Hungarian algorithm is used to solve the maximum weighted matching to ensure that the drones in each candidate task alliance meet the resource and time constraints. The optimal candidate task alliance is selected from all candidate task alliances, including: calculating the global utility change of each candidate task alliance; selecting the candidate task alliance with the largest global utility change from all global utility changes as the optimal candidate task alliance. Calculate the global utility change of each candidate task alliance: ΔSR = SR CS′ -SR CS残 , select the candidate task alliance with the largest ΔSR as the optimal candidate task alliance. Where CS′ represents the Nash equilibrium alliance structure including the optimal candidate task alliance after redistribution, CS 残 =CS|Task alliance affected by the failure.

[0121] After task reallocation, the system must undergo dynamic adaptive adjustments to ensure the entire task allocation system can adapt to the new post-failure situation. This phase primarily involves updating the coalition structure and reclaiming the resources of the faulty drone to avoid resource waste. Simultaneously, the optimal candidate task coalition is inserted into the Nash equilibrium coalition structure CS, and ineffective coalitions associated with the failure are removed, making the Nash equilibrium coalition structure CS more rational. Furthermore, to ensure the stability and rationality of task allocation, it is necessary to verify whether the Nash equilibrium coalition structure CS after reallocation satisfies Nash equilibrium conditions. Specifically, this involves verifying whether the utility change resulting from any drone's coalition shift satisfies the potential game conditions. If so, the Nash equilibrium coalition structure CS after reallocation is stable, and task allocation has reached an equilibrium state under the new circumstances, enabling sustained and efficient task execution. If not, further adjustments are required until the Nash equilibrium coalition structure CS satisfies the potential game Nash equilibrium conditions. This series of dynamic adaptive adjustments ensures that the system can quickly recover stability and continue to efficiently execute tasks in the face of emergencies such as drone failures.

[0122] The task reallocation mechanism in the embodiment of the present invention in the event of a fault adopts a dynamic task time window adjustment strategy to ensure that the tasks of the faulty drone can be quickly transferred to the healthy drone. The designed task splitting and reinsertion mechanism can ensure that task adjustments will not affect the completion of the original tasks.

[0123] For example, the heterogeneous drone task allocation and reallocation method based on alliance formation game proposed in an embodiment of the present invention can be applied to a dynamic delivery system for urban logistics drone swarms. Within a 1500m x 1500m downtown distribution area, five logistics sorting centers are located, responsible for the delivery of fresh produce, medical supplies, and general packages, respectively. Each center manages a 300m x 300m sub-area, and drones perform tasks within a 500m radius of the sorting center. Twenty heterogeneous drones with varying payload, temperature control, flight time, and communication capabilities are deployed within this area to fulfill the needs of 40 dynamic orders. These orders involve factors such as cargo weight, temperature control requirements, time constraints, and communication stability. In this scenario, heterogeneous drones can simultaneously perform different types of tasks, and their execution efficiency varies depending on the task type. These characteristics can be quantified as differences in the type and quantity of resources carried by the drones. Furthermore, each task consists of multiple subtasks, each with varying resource requirements. During task allocation, multiple drones can collaborate to perform the same task, forming a "drone alliance." Once the alliance arrives at the task location, the task is considered completed. In addition, the UAV may experience irreversible permanent failures during mission execution, resulting in the need to reallocate the mission. The method proposed in the present invention can achieve task reallocation.

[0124] In summary, the heterogeneous UAV task allocation and reallocation method based on alliance formation game proposed in the embodiment of the present invention, in order to improve the efficiency of task allocation, firstly, UAVs and tasks are intelligently grouped, and cluster centers are dynamically adjusted, thereby improving the matching accuracy of heterogeneous UAVs and tasks, reducing resource waste, and improving the overall task execution efficiency; by modeling an alliance formation game model, a distributed task allocation architecture is realized. This distributed task allocation architecture reduces communication and computing overhead and is more suitable for large-scale UAV cluster task scheduling compared to traditional centralized algorithms, enabling UAVs to collaboratively perform tasks, reducing task conflicts, and adaptively adjusting task allocation in a dynamic environment, thereby improving overall execution efficiency. By designing a benefit function, the contribution, path cost, and risk factors of UAVs are comprehensively considered to make task allocation more reasonable; combining the two-stage processing strategy of the maximum weighted matching stage and the UAV transfer alliance judgment stage, the computing efficiency is improved, ensuring that task allocation can be completed within a reasonable time; and designing a task reallocation mechanism to ensure that the tasks of faulty UAVs can be quickly transferred to healthy UAVs to avoid task interruption. In general, the embodiments of the present invention combine the complete process of clustering preprocessing, game theory decision-making, distributed task allocation and reallocation, ensuring that the task allocation scheme adapts to dynamic environmental changes and is suitable for multiple different application scenarios such as logistics and emergency rescue. It enhances the versatility and practicality of the task allocation scheme and has wider application value.

[0125] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0126] Although the present invention is described herein in conjunction with various embodiments, those skilled in the art may understand and implement other variations of the disclosed embodiments by reviewing the specification and accompanying drawings in the process of implementing the claimed invention. In the specification, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple components or steps. The fact that certain measures are described in different embodiments does not mean that these measures cannot be combined to produce good results.

[0127] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game, characterized by: The method comprises: Based on the drone status information, all drones are clustered to obtain a number of drone clusters, and based on the task status information, all tasks are clustered to obtain a number of task clusters, where the drone clusters correspond one to one to the task clusters; For each drone cluster and corresponding task cluster, the execution steps include: initializing the task alliance and modeling a coalition formation game model including a drone set, a task set, a payoff function for evaluating all drones in the task alliance, and evaluating the utility of individual drones in each task alliance; based on the coalition formation game model, constructing a payoff function for each drone in the task alliance, including a resource contribution function, a path cost function, and a threat cost function; based on the Nash equilibrium theory in game theory, modeling the task allocation problem as a task alliance partitioning problem that maximizes a potential function according to the payoff function; solving the task alliance partitioning problem that maximizes the potential function based on a two-stage process of a maximum weighted matching stage and a drone transfer alliance judgment stage, and obtaining a Nash equilibrium coalition structure to achieve drone task allocation; When a drone fails, the Nash equilibrium alliance structure of the failed drone is determined, and the failed drone is removed from the determined Nash equilibrium alliance structure, and the task corresponding to the failed drone is used as the task to be assigned; in the Nash equilibrium alliance structure after the failed drone is removed, the insertion positions of multiple tasks to be assigned are determined to form several candidate task alliances; based on the two-stage processing of the maximum weighted matching stage and the drone transfer alliance judgment stage, the task alliance partitioning problem of maximizing the potential function corresponding to several candidate task alliances is solved to select the optimal candidate task alliance from all candidate task alliances, and the optimal candidate task alliance is added to the Nash equilibrium alliance structure to realize the task reallocation of the failed drone.

2. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 1 is characterized in that: The drone status information includes drone resources, drone location, drone value, and drone speed; The mission status information includes the number of UAV resources required for the mission, the mission location, the mission hard time window information, and the mission threat level.

3. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 2 is characterized in that: Based on the drone status information, all drones are clustered to obtain several drone clusters, including: Divide all drones into several drone clusters; For each drone cluster, the execution steps include: calculating the first proportion of each drone in the drone cluster under each type of resource based on drone resources; calculating the resource entropy value of the corresponding drone based on all first proportions, selecting the drone position corresponding to the largest resource entropy value from all resource entropy values as the drone cluster center, calculating the first distance error between each drone in the drone cluster and the drone cluster center, judging whether the sum of all first distance errors meets the first preset threshold, if so, calculating the resource balance of the drone cluster, judging whether the resource balance of the drone cluster is less than or equal to the second preset threshold, if so, it indicates that each type of resource in the drone cluster meets the balance requirement, if not, returning to the step of dividing all drones into several drone clusters until each type of resource in all drone clusters meets the balance requirement.

4. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 2 is characterized in that: All tasks are clustered based on task status information to obtain several task clusters, including: Divide all tasks into several task clusters; the number of the task clusters is equal to the number of the drone clusters; For each task cluster, the execution steps include: calculating the second proportion of the number of each type of drone resources required for each task in the task cluster based on the number of drone resources required for the task to the total number of all types of drone resources required for the task; calculating the resource demand entropy value of the corresponding task based on all second proportions, selecting the task position corresponding to the largest resource demand entropy value from all resource demand entropy values as the task cluster center, calculating the second distance error between each task in the task cluster and the task cluster center, judging whether the sum of all second distance errors meets the third preset threshold, and if so, calculating the resource balance of the task cluster, judging whether the resource balance of the task cluster is less than or equal to the fourth preset threshold, and if so, indicating that each type of resource in the task cluster meets the balance requirement, and if not, returning to the step of dividing all tasks into several task clusters until each type of resource in all task clusters meets the balance requirement.

5. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 1 is characterized in that: The constructed profit function is expressed as follows: in, Indicates drone u i In the task j The profit under i ,t j ) indicates drone u i In the task j The resource contribution function under COST(u i ,t j ) indicates drone u i In the task j The path cost function under RISK(u i ,t j ) indicates drone u i In the task j The threat cost function under , α, β, γ all represent constant weight values; Resource contribution function VAL(u i ,t j ), the formula is: Among them, η, λ represent constant weights, RN represents the UAV u i Join the task j The number of resources available in the corresponding task alliance, Indicates drone u i Join the task j The number of unused resources in the corresponding task alliance; Path cost function COST(u i ,t j ), the formula is: Among them, d(u i ,t j ) indicates drone u i With the task t j The Euclidean distance between Represents task t j The size of the task environment area, μ represents a constant less than 0; Threat cost function RISK(u i ,t j ), the formula is: in, Indicates drone u i The value of Represents task t j threat level.

6. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 1 is characterized in that: Based on the two-stage process of maximum weighted matching and drone transfer alliance judgment, the task alliance partitioning problem of maximizing the potential function is solved and the Nash equilibrium alliance structure is obtained, including: In the maximum weighted matching phase, a bipartite graph is constructed for each task alliance, and maximum weighted matching is performed on the bipartite graph to obtain the corresponding initial allocation result. The construction of the bipartite graph for each task alliance includes: the task set and the drone set in each task alliance are used as the first set and the second set in the bipartite graph, and the maximum value of the benefit of each drone after joining the corresponding task alliance is used as the weight of the edge in the bipartite graph. In the drone transfer alliance judgment stage, the drone transfer alliance rules are designed. According to the initial allocation results and the drone transfer alliance rules, the drone transfer alliance is made to reach a stable state, and all drone transfer alliances are completed to obtain a Nash equilibrium alliance structure; the design of the drone transfer alliance rules includes adding the allocated drones in the initial allocation results to other task alliances for utility comparison. When the utility of the drone joining other task alliances is greater than the utility of the drone in the original task alliance, the drone exits the original task alliance and waits for reallocation. Otherwise, the drone has reached a stable state.

7. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 6 is characterized in that: The process of constructing a bipartite graph for each task alliance specifically includes: Mark all tasks in the task alliance as unassigned, all drones are not assigned tasks, and the drone alliance corresponding to each task is an empty set; Calculate the benefits of each UAV in the task alliance when performing each task individually to obtain the benefit matrix; The profit matrix is converted into a bipartite graph, and the bipartite graph is subjected to maximum weighted matching to obtain the corresponding initial allocation result. The conversion and matching process of the bipartite graph includes: taking each drone in the drone set as a vertex in the first set of the bipartite graph, and the value of each vertex in the first set is 0; taking each task in the task set as a vertex in the second set of the bipartite graph, and the value of each vertex in the second set is the maximum value of the profit of each drone after joining the task alliance individually; using the idea of the Hungarian algorithm to perform maximum weighted matching on the bipartite graph to obtain the initial allocation result.

8. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 7 is characterized in that: The process of designing the rules for the drone transfer alliance specifically includes: Calculate the first utility of the drone in the original mission alliance; Calculate the secondary utility of UAVs joining other mission coalitions; Compare whether the second utility is greater than the first utility. If so, the drone exits the original task alliance and waits for reallocation. If not, the drone has reached a stable state and remains in the original task alliance. Calculate the benefits of the remaining drones in the task alliance when they perform each remaining task individually to obtain a new benefit matrix. Return to the step of converting the benefit matrix into a bipartite graph and performing maximum weighted matching on the bipartite graph to obtain the corresponding initial allocation result until all drones reach a stable state.

9. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 1 is characterized in that: Determine the insertion positions of the plurality of tasks to be assigned, and form a plurality of candidate task alliances, including: Recalculate the earliest executable time of the task to be assigned based on the time when the faulty drone fails; According to the earliest executable time, the task execution time axis is divided into a predecessor task window, a successor task window and an independent time window; A plurality of insertion positions satisfying constraint conditions are determined within the predecessor task window, the successor task window, or the independent time window, and the tasks to be assigned are inserted into the insertion positions to obtain a plurality of candidate task alliances.

10. The method for allocating and reallocating heterogeneous UAV tasks based on alliance formation game according to claim 1 is characterized in that: Select the best candidate task alliance from all candidate task alliances, including: Calculate the global utility change of each candidate task alliance; The candidate task alliance corresponding to the largest global utility change is selected from all global utility changes as the optimal candidate task alliance.

Images