Heterogeneous unmanned aerial vehicle task allocation method based on task coalition and individual utility

Abstract

The application discloses a heterogeneous unmanned aerial vehicle task allocation method based on task alliance and individual utility, first constructs a heterogeneous cluster containing attack unmanned aerial vehicles and reconnaissance unmanned aerial vehicles and a task set model, sets an optimal alliance size and a constraint condition; secondly, establishes an utility evaluation rule based on the alliance size, quantifies individual utility and alliance residual value; subsequently, the attack unmanned aerial vehicles autonomously select an alliance according to an individual utility maximization principle, and eliminate conflicts through local interaction; finally, the reconnaissance unmanned aerial vehicles make two-dimensional utility decision and iterative adjustment according to residual utility demand of the task alliance and self cost, so that accurate matching and efficient allocation of heterogeneous resources to tasks are realized.

Description

Technical Field

[0001] This invention relates to the technical field of unmanned aerial vehicle (UAV) swarm control, and in particular to a heterogeneous UAV task allocation method based on task alliances and individual utility. Background Technology

[0002] As unmanned systems technology evolves towards intelligence and swarming, heterogeneous UAV swarm collaborative operations have become a key means of executing complex tasks. In practical applications, heterogeneous swarms composed of attack and reconnaissance UAVs often need to collaboratively complete reconnaissance and precision strike missions against high-value targets in complex environments with strong electromagnetic interference, limited communication bandwidth, and even intermittent link interruptions. How to achieve efficient task allocation and resource scheduling under such constrained conditions has become a core issue in improving the overall capability of the swarm.

[0003] Existing task allocation methods are mainly divided into two categories: centralized and distributed. Centralized methods rely on ground stations or a central master unit for global coordination. While they can theoretically obtain the globally optimal solution, they suffer from a serious risk of single point of failure. Furthermore, as the cluster size increases, the transmission of massive amounts of status information can lead to communication congestion, making it difficult to meet the high real-time requirements of the battlefield. In contrast, distributed methods endow individual UAVs with autonomous decision-making capabilities, requiring only local interactions between neighboring nodes to complete task allocation. This significantly reduces communication load and provides stronger robustness and scalability.

[0004] However, when dealing with heterogeneous drone swarms composed of drones with different functions such as attack and reconnaissance, existing distributed methods, such as traditional contract networks, consensus packet algorithms, or single auction algorithms, still have significant limitations. For example, most existing technologies are not adapted to heterogeneous characteristics, often employing a single allocation mechanism to homogenize all nodes, ignoring the functional differences and dependencies between different types of drones. Specifically, existing solutions struggle to achieve efficient tactical coordination when handling heterogeneous swarms, and resource utilization cannot reach its optimal level.

[0005] To address the aforementioned problems, this invention proposes a hierarchical distributed task allocation method for heterogeneous UAV swarms. This method introduces a phased decision-making mechanism, based on task alliances and individual utility, effectively resolving coordination conflicts between heterogeneous nodes and achieving rapid convergence and efficient tactical coordination in swarm task allocation under complex adversarial environments. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings and deficiencies of the prior art and provide a heterogeneous UAV task allocation method based on task alliances and individual utility, so as to achieve accurate matching and efficient allocation of heterogeneous resources to tasks.

[0007] To achieve the above objectives, the technical solution provided by this invention is: a heterogeneous UAV task allocation method based on task alliances and individual utility, comprising the following steps:

[0008] S1: Digital modeling of heterogeneous UAV swarms and task sets; the heterogeneous UAV swarm includes two types of UAVs: attack UAVs and reconnaissance UAVs; the task set includes selectable target tasks and an empty task to accommodate unassigned UAVs; UAVs that select the same task are defined as a task alliance, and an optimal alliance size is set for each task alliance, the optimal alliance size being the number of UAVs required for the task alliance to obtain maximum utility; constraints are also set, including single-UAV task uniqueness constraints, alliance size limit constraints, and reconnaissance resource constraints.

[0009] S2: After modeling is completed, construct the task alliance utility evaluation rules and individual utility calculation rules. First, define the utility function of the task alliance, which takes the number of UAVs in the task alliance as the independent variable and exhibits a characteristic of first increasing and then decreasing, reaching its maximum value when the optimal alliance size is reached, thus representing the task alliance's demand for UAV size. Second, construct the individual utility calculation rules for attack UAVs, that is, define the incremental utility of the task alliance caused by adding attack UAVs to the task alliance as the individual utility of the attack UAVs. Third, after the allocation of attack UAVs is determined, the remaining utility of the task alliance is defined as the task alliance's utility demand for reconnaissance UAVs. Finally, define the individual costs of the two types of UAVs respectively, thus representing the matching degree between the two types of UAVs and different tasks.

[0010] S3: Each attack drone calculates the utility of joining each task alliance based on the individual utility and cost function defined in step S2, and autonomously selects the task alliance with the highest utility to join; then, the attack drones exchange local information. If a selection conflict occurs, a conflict resolution rule is constructed using the number of iterations and the cost function, that is, priority is given to retaining the task selection with the higher number of iterations or the lower cost. The party that fails to allocate the task will reselect the task until the cluster reaches a stable state without conflict, completing the initial construction of the task alliance. Finally, the attack drones broadcast the task alliance information to the reconnaissance drones.

[0011] S4: Based on the task alliances determined in step S3, calculate the utility requirement of each task alliance for reconnaissance drones. The utility requirement represents the degree of demand for reconnaissance drones by the task alliance at the current scale. Each reconnaissance drone constructs a comprehensive utility index that includes the utility requirement of the task alliance and the individual cost. The reconnaissance drones make iterative decisions based on the comprehensive utility index and autonomously choose to join the task alliance with the highest utility. When multiple reconnaissance drones choose the same task alliance, resulting in a conflict, the matching relationship between the reconnaissance drone with the highest utility and the task alliance is determined according to the principle of maximum utility priority. The remaining reconnaissance drones enter the next round of iteration and continue to choose to join the task alliance with the highest utility from the remaining task alliances. Through multiple rounds of iteration until the selection of all drones is determined, the final heterogeneous drone cluster task allocation result is output.

[0012] Furthermore, the specific steps of step S1 are as follows:

[0013] S11: Digitally model the heterogeneous drone swarm, which includes... attack drones and Reconnaissance drones ,in and They represent the first attack drones and the first A reconnaissance UAV; defining the state information of heterogeneous UAVs, including position vectors. and velocity vector and Based on the positional relationships between drones, a communication connectivity matrix is ​​established to determine the communication topology within the cluster, and all drones that can communicate with each other are referred to as neighboring drones.

[0014] S12: Define all enemy aircraft within the detection range of the heterogeneous drone swarm as tasks to be assigned, and construct a system containing... A set of tasks ,in Indicates the first One task, Indicates an empty task; each task The state information includes the position vector. and velocity vector Because tasks also have heterogeneous characteristics, each task is configured... The optimal alliance size for the corresponding mission alliance is For air-to-air missions, there are no restrictions on the size of the alliance;

[0015] S13: Define binary decision variables and , respectively representing attack drones and reconnaissance drones Select task? If selected, the variable value is 1; otherwise, it is 0. Based on actual needs, define the constraint rules that task allocation must meet, including:

[0016] Decision variable constraints:

[0017] ;

[0018] Single-machine task uniqueness constraint: Any drone can only select one task to be assigned or an empty task at any given time, that is, satisfying:

[0019] ;

[0020] Alliance size constraint: Execute any task The total number of drones shall not exceed the optimal coalition size set for this mission. That is, satisfying:

[0021] ;

[0022] Reconnaissance resource constraints: In a context where any mission In the alliance formed, the number of reconnaissance drones does not exceed one, that is, the following condition is met:

[0023] .

[0024] Furthermore, the specific steps of step S2 are as follows:

[0025] S21: Construct task alliance utility evaluation rules, i.e., for tasks The formed task alliance Define its task alliance utility function for:

[0026] ;

[0027] In the formula: Indicates joining the mission alliance. The number of drones, It is for the task The best league size, Indicates when exactly A drone joins the mission coalition The maximum utility that the alliance can obtain in this mission;

[0028] S22: Define the individual utility of an attack drone, which is defined as the difference between the utility of the mission coalition after the attack drone joins the coalition and before joining the coalition. Therefore, the individual utility of an attack drone is... Join the Mission Alliance The utility for:

[0029] ;

[0030] In the formula: It is an attack drone The benefits of joining a mission alliance. This is the current utility of the Mission Alliance;

[0031] S23: Defines the utility requirement of a mission coalition for a reconnaissance UAV, which is defined as the difference between the maximum utility of the mission coalition and the utility of the reconnaissance UAV before joining the mission coalition. For reconnaissance drones utility needs for:

[0032] ;

[0033] S24: Construct the individual cost calculation function for the attack drone, comprehensively considering distance and heading factors, and define the attack drone. Join the Mission Alliance Individual cost for:

[0034] ;

[0035] In the formula: It is the distance correlation coefficient. It is the velocity correlation coefficient. This represents the Euclidean distance between the attack drone and the mission target. It is the cosine angle between the velocity vector of the attack drone and the velocity vector of the mission target;

[0036] S25: Construct the individual cost calculation function for reconnaissance UAVs, and define the reconnaissance UAV. Join the Mission Alliance Individual cost for:

[0037] ;

[0038] In the formula: This represents the Euclidean distance between the reconnaissance drone and the mission target. It is the optimal distance that reconnaissance drones and mission targets are expected to maintain. It is the cosine angle between the velocity vector of the reconnaissance UAV and the velocity vector of the mission target.

[0039] Furthermore, the specific steps of step S3 are as follows:

[0040] S31: Definition and initialization of attack drone state variables: Define the satisfaction state variable `satisfied` for the attack drone, which is a binary variable used to indicate whether the attack drone has found a satisfactory expected task; define the iteration count `Iter`, which is an integer variable used to compare the iteration counts of the attack drones to resolve possible selection conflicts during algorithm execution; define the self-alliance cognition `CS`, which is a set used to store the current task assignment status of all attack drones from the perspective of this attack drone;

[0041] Before the algorithm starts, the above variables are initialized: satisfied is set to false, indicating that the currently selected task is not yet satisfactory; Iter is set to 0; CS is initialized to S0, that is, by default all attack drones select an empty task at the initial moment.

[0042] S32: Determine the satisfaction state variable of the attack drone. If it is true, maintain the current state; if it is false, proceed to the mission selection phase. This phase involves iterating through all missions and calculating the utility of joining each mission coalition. for:

[0043] ;

[0044] All drones that choose an empty mission form an empty mission alliance. The utility of joining the empty mission alliance is defined as fixed at 0. Then, the drone with the highest utility is selected to join its mission alliance. At the same time, the iteration count Iter is incremented by 1, the satisfaction state variable satisfied is updated to true, and the alliance cognition CS is updated according to its own mission selection scheme.

[0045] S33: The attack drone broadcasts its own alliance cognition CS and iteration number Iter to all neighboring attack drones within the communication topology. At the same time, it receives the alliance cognition and iteration number information broadcast by neighboring attack drones, and combines the received information with its own information to form a list of "iteration number - alliance cognition" information pairs to be processed.

[0046] S34: Traverse the list of "iteration count - alliance cognition" information pairs obtained in step S33, and update the decision according to the following rules:

[0047] Prioritize adopting alliance cognitive information with a larger iteration count (Iter);

[0048] When the number of iterations is the same, compare the costs of the attack drone in the corresponding mission coalition. Prioritize adopting alliance cognitive information that has lower costs;

[0049] If the neighbor's information is adopted according to the above rules, that is, if the neighbor's information is found to be better than its own information, then the neighbor's information is used to update its own "iteration count - alliance cognition" information to modify its own alliance cognition and iteration count, and its own satisfaction state variable satisfied is reset to false to trigger the next round of task selection; otherwise, no modification is made.

[0050] S35: Repeat steps S32 to S34 until the satisfaction status variable satisfied of all attack drones in the cluster is true, which indicates that the task allocation at the attack drone level is complete.

[0051] S36: After the attack drone completes the mission assignment, it will broadcast the final mission coalition to the reconnaissance drone.

[0052] Furthermore, the specific steps of step S4 are as follows:

[0053] S41: Based on the task alliance generation results in step S36 and the calculation formula in step S23, calculate the utility requirement of each task alliance for the reconnaissance UAV. And combine all utility demands into a task alliance bid list. ,set up Indicates joining the mission alliance For reconnaissance drones Utility needs;

[0054] S42: Calculate the reconnaissance UAV's entry into the mission coalition based on the formula in step S25. The cost And all the cost combinations are defined as the cost list for reconnaissance drones. ,set up Reconnaissance drone Join the Mission Alliance The cost;

[0055] S43: Create a task assignment list and the list of maximum utility ,set up Reconnaissance drone Choose to join the mission alliance , This indicates the utility of a reconnaissance UAV in selecting this mission coalition; initialize each value in both lists to 0.

[0056] S44: Traverse the task assignment list The task is to calculate the utility of selecting this task. :

[0057] ;

[0058] Finally, select the task with the highest utility as the current target task and update its task allocation list accordingly. and the list of maximum utility ;

[0059] S45: Sends its own mission assignment list to all neighboring reconnaissance drones. and the list of maximum utility It also receives the task assignment list and maximum utility list of all neighboring reconnaissance drones;

[0060] S46: Traverse the task allocation list of neighbor reconnaissance drones and check for task selection conflicts; if a neighbor reconnaissance drone is found to be more effective for the same task, then the neighbor reconnaissance drone is determined to win and abandons its selection of that task, and the task allocation list is updated based on the information from the neighbor reconnaissance drone. and the list of maximum utility To resolve the conflict;

[0061] S47: Check if the task allocation list of the drone itself has changed after step S45. If it has changed, it means that the task has been taken over by other reconnaissance drones. Continue to execute step S44 to reselect the task. If the task allocation lists of all reconnaissance drones have not changed, the task allocation of the heterogeneous drone cluster is completed.

[0062] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0063] 1. This invention abandons the traditional approach to heterogeneous UAV task allocation and proposes a phased decision-making task allocation architecture. By first allocating attack-type UAV tasks, and then calculating the utility requirements of the task alliance for reconnaissance-type UAVs based on a determined task alliance, the originally complex global decision-making problem of heterogeneous clusters is decoupled into a phased serial decision-making problem. This strategy effectively compresses the search space and significantly reduces the time complexity of the algorithm while ensuring the quality of task allocation results.

[0064] 2. This invention constructs utility evaluation rules for two types of UAVs respectively. On the one hand, by introducing a mission alliance utility function with the characteristic of first increasing and then decreasing, combined with the optimal alliance size constraint, it ensures that attack UAVs will not be over-concentrated when autonomously forming mission alliances. On the other hand, by utilizing the utility requirements of mission alliances for reconnaissance UAVs, the complementary relationship between attack UAVs and reconnaissance UAVs is quantified. Combined with the individual cost of adapting the two types of UAVs, it ensures that each UAV can join the most needed mission alliance, maximizing the collaborative combat capability of heterogeneous clusters.

[0065] 3. This invention employs a distributed information exchange mechanism, eliminating the need for a central node for global control. Similar drones only need to exchange information between neighboring nodes, and two types of drones only need to exchange information once after the attack drone's mission allocation is completed. This mechanism significantly reduces the data throughput and communication frequency required during mission allocation, alleviating bandwidth pressure and preventing single-point failures from impacting the entire cluster.

[0066] 4. This invention, by constructing explicit conflict resolution rules, enables UAVs to have a clear decision-making orientation when facing task selection conflicts, and can quickly converge to a stable state with fewer iterations. Furthermore, this method has good dynamic adaptability; when the cluster size changes, it can adaptively adjust the allocation scheme without reconstructing the overall architecture, thus meeting the needs of task allocation in large-scale heterogeneous clusters. Attached Figure Description

[0067] Figure 1 This is a decision-making flowchart for attack drones.

[0068] Figure 2 This is a decision-making flowchart for reconnaissance drones.

[0069] Figure 3 This is a schematic diagram showing the result of task allocation in the embodiment.

[0070] Figure 4 This is a utility curve of task allocation in the implementation example.

[0071] Figure 5 This is a diagram showing the task assignment results for a newly added task.

[0072] Figure 6 This is a diagram illustrating the task allocation results after deleting a task. Detailed Implementation

[0073] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0074] This embodiment discloses a heterogeneous UAV task allocation method based on task alliances and individual utility. This method reduces the computational pressure of traditional heterogeneous cluster task allocation, decouples the complex heterogeneous cluster coordination problem into a phased evolution process of sub-clusters, and achieves effective dimensionality reduction of the high-dimensional solution space by locking the task objectives and resource states in stages, thus providing support for the autonomous decision-making of the cluster in adversarial environments. The specific details are as follows:

[0075] Step 1: Digitally model the heterogeneous drone cluster and task set.

[0076] First, a digital model of the heterogeneous drone swarm is performed, the heterogeneous drone swarm comprising... attack drones and Reconnaissance drones ,in and They represent the first attack drones and the first A reconnaissance drone, the state information of which includes a position vector. and velocity vector and The model establishes a communication connectivity matrix based on the positional relationships between UAVs to determine the communication topology within the cluster, defining all UAVs capable of communicating with each other as neighboring UAVs. Simultaneously, the model mandates that the cluster topology remain fully connected, meaning there must be an information transmission path between any two nodes. In practical control, a virtual clustering force is introduced as a constraint; when the distance between UAVs tends to increase significantly, an attractive force is generated, thus maintaining the cluster's communication connectivity during dynamic maneuvers. All enemy aircraft within the detection range of the heterogeneous UAV cluster are defined as mission targets to be assigned, and a communication topology matrix is ​​constructed containing... A set of tasks ,in Indicates the first One task, Indicates an empty task; each task The state information includes the position vector. and velocity vector Because tasks also have heterogeneous characteristics, each task is configured... The optimal alliance size for the corresponding mission alliance is Empty missions do not limit the size of the alliance.

[0077] Define binary decision variables and , respectively representing attack drones and reconnaissance type Select task? If selected, the variable value is 1; otherwise, it is 0. Based on actual needs, define the constraint rules that task allocation must meet, including:

[0078] Decision variable constraints:

[0079] ;

[0080] Single-machine task uniqueness constraint: Any drone can only select one task to be assigned or an empty task at any given time, that is, satisfying:

[0081] ;

[0082] Alliance size constraint: Execute any task The total number of drones shall not exceed the ideal coalition size set for this mission. That is, satisfying:

[0083] ;

[0084] Reconnaissance resource constraints: In a context where any mission In the alliance formed, the number of reconnaissance drones does not exceed one, that is, the following condition is met:

[0085] ;

[0086] Step 2: Construct the task alliance utility evaluation rules and individual utility calculation rules.

[0087] After modeling is completed, in order to drive the two types of drones to make autonomous decisions, it is necessary to construct a set of evaluation rules based on utility maximization. In the scenario of heterogeneous drone swarm cooperative combat, the ultimate goal of task allocation is to maximize group utility, a process that depends on the interaction between task alliance utility and individual utility.

[0088] A mission coalition, as a basic unit of cooperation, refers to a group of drones that choose the same task, integrating the local capabilities of individuals into a collective adversarial advantage. However, limited by communication and perception constraints, individual drones cannot grasp global information and can only assess their own individual utility based on current local information. Therefore, individual utility becomes the core driving force for decision-making. Drones continuously seek out and join coalitions that maximize their individual utility, thereby increasing both their own utility and the overall utility of the group. When all drones have maximized their individual utility, and no drone can achieve greater efficiency by unilaterally altering the mission coalition, the mission coalition converges to a stable state, thus achieving global task allocation optimization under local perception conditions.

[0089] To achieve a local advantage of "many to one" while preventing resource waste—specifically, preventing too many drones from attacking the same target and causing utility overflow—a task alliance utility evaluation rule is constructed, targeting specific tasks. Mission Alliance Define its task alliance utility function for:

[0090] ;

[0091] In the formula: Indicates joining the alliance for this mission. The number of drones, It is for the task The best league size, Indicates when exactly A drone joins the mission coalition This represents the maximum utility that the task alliance can obtain. The function has a property of first increasing and then decreasing, when... At that time, the addition of new members significantly increases utility, thereby guiding drones to join the formation of alliances; when In some cases, the addition of new members can actually lead to a decrease in utility, thereby inhibiting overcrowding.

[0092] The individual utility of an attack drone is defined as the difference between its utility after joining the mission coalition and its utility before joining, since it is expected that multiple attack drones joining a mission coalition will create a local advantage. Join the alliance Individual utility for:

[0093] ;

[0094] In the formula: It is an attack drone The benefits of joining a mission alliance This is the current utility of the mission alliance.

[0095] The utility requirement of a mission coalition for reconnaissance drones is defined as follows: Since reconnaissance drones are expected to play a reconnaissance role and guide attack drones performing air missions into combat, the fewer friendly attack drones in the current mission coalition, the stronger their motivation to intervene. Simultaneously, to prevent resource waste caused by multiple reconnaissance drones in one mission coalition, this utility requirement is defined as the difference between the maximum utility of the mission coalition and the utility of the reconnaissance drone before it joins the coalition. That is, once a reconnaissance drone joins the mission coalition, the coalition will no longer have any remaining utility to provide to other drones. Therefore, the mission coalition... For reconnaissance drones utility needs for:

[0096] ;

[0097] Whether the two types of drones decide to join the mission alliance depends not only on their individual utility or utility needs, but also on their mission preferences. To reflect the tactical characteristics of the two types of drones, differentiated individual cost calculation functions are set up as deductions for calculating net utility.

[0098] Attack drones prioritize engaging nearby targets to minimize range loss and gain a first-strike advantage. They also favor targets with relatively opposing velocity vectors, meaning they prioritize rapidly approaching enemy threats. The individual cost of an attack drone is constructed by comprehensively considering distance and heading factors, defining the attack drone as such. Join the Mission Alliance The cost for:

[0099] ;

[0100] In the formula: It is the distance correlation coefficient. It is the velocity correlation coefficient. This represents the Euclidean distance between the attack drone and the mission target. It is the cosine angle between the velocity vector of the attack drone and the velocity vector of the mission target.

[0101] Preferred reconnaissance drones and current optimal reconnaissance distance Approaching enemy aircraft prefer targets with similar velocity vectors to monitor targets that pose little threat now but may pose a future threat or are escaping. A cost function for the reconnaissance UAV is constructed, and the reconnaissance UAV is defined. Join the Mission Alliance The cost for:

[0102] ;

[0103] In the formula: This represents the Euclidean distance between the reconnaissance drone and the mission target. It is the optimal distance that reconnaissance drones and mission targets are expected to maintain. It is the cosine angle between the velocity vector of the reconnaissance UAV and the velocity vector of the mission target.

[0104] Step 3: Formation of an alliance of attack drones.

[0105] For a single attack drone, the specific process of mission allocation is as follows: Figure 1As shown in the diagram, when the task allocation algorithm starts, the drone first initializes its key state variables: It sets the satisfaction state variable `satisfied` to `false`, indicating that the drone believes the current task allocation scheme may not be optimal and requires further calculation; `true` indicates satisfaction with the current scheme. It sets the iteration count `Iter` to 0, a logical counter variable used to determine which attack drones have the most up-to-date alliance covenant knowledge to resolve conflicts in distributed decision-making. It initializes the current alliance knowledge `CS` to `S0`, meaning that all attack drones initially select an empty task. After parameter initialization, the algorithm enters a loop. The convergence condition for exiting the loop is that during communication, the satisfaction state variable `satisfied` of all attack drones is found to be `true`, meaning no attack drone joining another task alliance will generate greater utility than it does in its original task alliance.

[0106] This loop comprises two core phases: task selection and conflict resolution. In the task selection phase, if the user is dissatisfied with the current task (satisfied = false), it enters the phase of finding the optimal task. The attack drone calculates the utility and cost C of joining all available task alliances based on the formula in step two, searching for the alliance that provides the greatest utility improvement. If the calculated utility of joining another alliance is greater than the current alliance, the drone changes its task alliance, increments the iteration count Iter by 1, updates its alliance awareness CS according to this scheme, and records its cost C; otherwise, no changes are made. Finally, satisfied is set to true, meaning that the drone believes it has made the optimal decision until it receives new external information.

[0107] Phase two is the conflict resolution phase. Regardless of whether the drones updated their missions in the previous phase, each attack drone broadcasts its own alliance cognition (CS) and iteration count (Iter) to all neighboring attack drones within the communication topology. Simultaneously, it receives alliance cognition and iteration count information broadcast by its neighbors, combining the received information with its own to form a list of pending "iteration count - alliance cognition" information pairs. Decision updates are then made according to the following rules:

[0108] Prioritize adopting alliance cognitive information with a larger iteration count (Iter);

[0109] When the number of iterations is the same, compare the costs of the attack drone in the corresponding mission coalition. Prioritize adopting alliance cognitive information that has lower costs;

[0110] If traversing the "iteration count - coalition awareness" information pair list according to the above rules, the maximum iteration count Iter_k, the corresponding coalition awareness CS_k, and the corresponding cost C_k are found. If it is found that Iter_k > Iter or Iter_k = Iter and C_k < C, that is, it is found that the coalition division of the neighbor is better than its own coalition division, then update its own "iteration count - coalition awareness" information pair with the coalition division of the neighbor, update its own iteration count variable Iter to the neighbor's Iter_k, update its own coalition awareness CS to the neighbor's CS_k, and reset its own satisfaction status variable satisfied to false to trigger the next round of re-evaluation. This is because once the coalition assignment plan of the neighbor is received, it means that the previous task coalition selection has failed. The current attack UAV must recalculate the coalition it is most suitable to join based on the currently determined coalition plan in the next loop until the system converges; otherwise, no modification is made.

[0111] Finally, after meeting the convergence condition for exiting the loop, the task assignment of the attack UAV is completed, and the attack UAV broadcasts the finally generated task coalition to the reconnaissance UAV.

[0112] Step 4: Coalition generation of the reconnaissance UAV.

[0113] After the attack UAV forms a stable coalition division, it enters the reconnaissance UAV task assignment stage. This stage adopts a distributed bidding auction mechanism based on the two-dimensional utility index of the utility demand and individual cost of the task coalition, aiming to allocate reconnaissance resources to the task coalition that most needs support.

[0114] As Figure 2 shown, first calculate the utility demand of each task coalition for the reconnaissance UAV , and combine all the utility demands, which is defined as the task coalition bid list , set to represent the utility demand of joining task coalition for the reconnaissance UAV , ensuring that the reconnaissance resources can flow to the task with the highest utility. Secondly, calculate the cost of the reconnaissance UAV joining task coalition , and combine all the costs, which is defined as the cost list of the reconnaissance UAV , set to represent the cost of the reconnaissance UAV joining task coalition .

[0115] Subsequently, enter the specific task assignment process of the reconnaissance UAV. Establish a task assignment list and the maximum utility list ,set up Reconnaissance drone Choose to join the mission alliance , This indicates the utility of the reconnaissance UAV in selecting this mission coalition. Initialize each value in both lists to 0;

[0116] After initialization, the algorithm enters an iterative process. All drones without assigned targets enter the bidding phase, traversing the task list. The task is to calculate the utility of selecting this task. :

[0117] ;

[0118] Finally, select the task with the highest utility as the current target task and update its task allocation list accordingly. and the list of maximum utility ;

[0119] After the bidding phase, each drone selects the task with the highest utility. However, to prevent duplicate selections, a consensus phase is needed to resolve conflicts. During the consensus phase, each drone sends its task allocation list to all neighboring reconnaissance drones. and the list of maximum utility It receives the task lists and maximum utility lists of all neighboring reconnaissance drones, iterates through the lists of neighboring reconnaissance drones, and checks for task selection conflicts. If the utility of its own selected task is not greater than that of any neighboring drones that have selected the same task, meaning a neighboring reconnaissance drone has a higher utility for the same task, then the neighboring drone is determined to win, and the drone itself must relinquish its selection of that task, remove itself from the task allocation list, and remove the corresponding utility from its maximum utility list. Based on the information from the neighboring reconnaissance drones, the maximum utility of the current task and the bidders are recorded in its own maximum utility list. and task assignment list The conflict is resolved. At this point, the drone that lost its target due to the competition will return to the unassigned state and reselect the optimal target from the remaining available tasks in the next iteration. This process continues until the task assignment lists of all reconnaissance drones remain unchanged, at which point the task assignment of the heterogeneous drone swarm is complete.

[0120] Figure 3This diagram illustrates the final convergence state of heterogeneous UAV swarm task allocation after 26 iterations. Six enemy targets, labeled T1-T6, are represented by diamond icons. Among the friendly heterogeneous UAVs, attack UAVs are represented by circular icons, and reconnaissance UAVs by square icons. Independent nodes not connected to other nodes represent idle UAVs not in the task alliance; UAVs in the task alliance are connected to their respective target nodes by solid lines, clearly demonstrating their affiliation. This invention achieves differentiated and precise allocation for different types of task requirements: targets T3 and T5, being high-demand targets, are each successfully allocated 4 UAVs for coverage, while other regular targets are allocated 3 UAVs as needed, perfectly matching the resource configuration of the heterogeneous swarm with task requirements. Furthermore, the entire allocation process reaches stability in just 26 iterations, indicating that this invention not only effectively solves the multi-target collaborative allocation problem of heterogeneous UAV swarms and accurately meets various tactical constraints, but also exhibits excellent convergence speed and execution efficiency in terms of computational performance.

[0121] Figure 4 The graph illustrates the evolution of the total utility of a heterogeneous drone swarm as a function of iterations during task allocation. The graph clearly shows a significant monotonically increasing trend in total utility with increasing iteration count, validating the effectiveness of this invention in global optimization. In the early stages of iteration, the curve slope is steep, indicating rapid utility growth. This is because the alliances are not yet saturated in the initial phase, and the utility of drones joining task alliances increases significantly, driving a rapid increase in global utility. As iterations progress, each alliance gradually approaches its preset resource requirement limit, the task allocation structure becomes more stable, leading to diminishing marginal returns and a slowdown in the utility growth rate, eventually entering a plateau.

[0122] Figure 5 The results demonstrate the dynamic redistribution of resources under a simulated scenario with a sudden addition of new tasks. The heterogeneous cluster reached a stable state in the 29th iteration. Compared to the initial scheme, a new target T7 with an optimal coalition size of 3 was suddenly added to the scenario. This dynamic change triggered the global response mechanism of this invention, prompting a large number of drones to flexibly adjust their original mission orientations based on the real-time situation. Particularly for target T1, due to the diversion of some resources to cope with the new task, a local shortage of its attack drones occurred. The algorithm demonstrated excellent heterogeneous collaborative capabilities, quickly dispatching a reconnaissance drone to continuously monitor T1 and guiding previously idle attack drones to maneuver towards the target to supplement execution power.

[0123] Figure 6This demonstrates the dynamic task reallocation results after the removal of some targets in the simulated environment, at which point the heterogeneous cluster reaches a stable state in the 25th iteration. Compared to the initial scheme, target T6, which originally had an optimal coalition size of 3, has been successfully removed, freeing up previously strained system resources. Due to the reduction in the total number of tasks, ample execution resources have become available in some areas. Figure 6 As shown, the remaining targets T1-T5 can be efficiently completed using only attack drones, without the need for complex heterogeneous collaboration to fill gaps. This fully demonstrates the high flexibility and excellent coordination mechanism of this invention under different mission situations, enabling intelligent complementarity when resources are scarce and automatic configuration optimization when resources are abundant.

[0124] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A heterogeneous UAV task allocation method based on task alliances and individual utility, characterized in that, Includes the following steps: S1: Digital modeling of heterogeneous UAV swarms and task sets; the heterogeneous UAV swarm includes two types of UAVs: attack UAVs and reconnaissance UAVs; the task set includes selectable target tasks and an empty task to accommodate unassigned UAVs; UAVs that select the same task are defined as a task alliance, and an optimal alliance size is set for each task alliance, the optimal alliance size being the number of UAVs required for the task alliance to obtain maximum utility; constraints are also set, including single-UAV task uniqueness constraints, alliance size limit constraints, and reconnaissance resource constraints. S2: After modeling is completed, construct the task alliance utility evaluation rules and individual utility calculation rules. First, define the utility function of the task alliance. This utility function takes the number of drones in the task alliance as the independent variable and exhibits the characteristic of first increasing and then decreasing. It reaches its maximum value when the optimal alliance size is reached, thereby representing the task alliance's demand for the number of drones. Second, construct the individual utility calculation rules for attack drones. That is, the increase in task alliance utility generated by adding attack drones to the task alliance is defined as the individual utility of the attack drones. Secondly, after the allocation of attack drones is determined, the remaining utility of the mission coalition is defined as the utility requirement of the mission coalition for reconnaissance drones; finally, the individual costs of the two types of drones are defined respectively to characterize the matching degree of the two types of drones with different missions. S3: Each attack drone calculates the utility of joining each task alliance based on the individual utility and cost function defined in step S2, and autonomously selects the task alliance with the highest utility to join; then, the attack drones exchange local information. If a selection conflict occurs, a conflict resolution rule is constructed using the number of iterations and the cost function, that is, priority is given to retaining the task selection with the higher number of iterations or the lower cost. The party that fails to allocate the task will reselect the task until the cluster reaches a stable state without conflict, completing the initial construction of the task alliance. Finally, the attack drones broadcast the task alliance information to the reconnaissance drones. S4: Based on the task alliances determined in step S3, calculate the utility requirement of each task alliance for reconnaissance drones. The utility requirement represents the degree of demand for reconnaissance drones by the task alliance at the current scale. Each reconnaissance drone constructs a comprehensive utility index that includes the utility requirement of the task alliance and the individual cost. The reconnaissance drones make iterative decisions based on the comprehensive utility index and autonomously choose to join the task alliance with the highest utility. When multiple reconnaissance drones choose the same task alliance, resulting in a conflict, the matching relationship between the reconnaissance drone with the highest utility and the task alliance is determined according to the principle of maximum utility priority. The remaining reconnaissance drones enter the next round of iteration and continue to choose to join the task alliance with the highest utility from the remaining task alliances. Through multiple rounds of iteration until the selection of all drones is determined, the final heterogeneous drone cluster task allocation result is output.

2. The heterogeneous UAV task allocation method based on task alliances and individual utility according to claim 1, characterized in that, The specific steps for step S1 are as follows: S11: Digitally model the heterogeneous drone swarm, which includes... attack drones and Reconnaissance drones ,in and They represent the first attack drones and the first A reconnaissance UAV; defining the state information of heterogeneous UAVs, including position vectors. and velocity vector and Based on the positional relationships between drones, a communication connectivity matrix is ​​established to determine the communication topology within the cluster, and all drones that can communicate with each other are referred to as neighboring drones. S12: Define all enemy aircraft within the detection range of the heterogeneous drone swarm as tasks to be assigned, and construct a system containing... A set of tasks ,in Indicates the first One task, Indicates an empty task; each task The state information includes the position vector. and velocity vector Because tasks also have heterogeneous characteristics, each task is configured... The optimal alliance size for the corresponding mission alliance is For air-to-air missions, there are no restrictions on the size of the alliance; S13: Define binary decision variables and , respectively representing attack drones and reconnaissance drones Select task? If selected, the variable value is 1; otherwise, it is 0. Based on actual needs, establish the constraints that task allocation must meet, including: Decision variable constraints: ； Single-machine task uniqueness constraint: Any drone can only select one task to be assigned or an empty task at any given time, that is, satisfying: ； Alliance size constraint: Execute any task The total number of drones shall not exceed the optimal coalition size set for this mission. That is, satisfying: ； Reconnaissance resource constraints: In a context where any mission In the alliance formed, the number of reconnaissance drones does not exceed one, that is, the following condition is met: 。 3. The heterogeneous UAV task allocation method based on task alliances and individual utility according to claim 2, characterized in that, The specific steps for step S2 are as follows: S21: Construct task alliance utility evaluation rules, i.e., for tasks The formed task alliance Define its task alliance utility function for: ； In the formula: Indicates joining the mission alliance. The number of drones, It is for the task The best league size, Indicates when exactly A drone joins the mission coalition The maximum utility that the alliance can obtain in this mission; S22: Define the individual utility of an attack drone, which is defined as the difference between the utility of the mission coalition after the attack drone joins the coalition and before joining the coalition. Therefore, the individual utility of an attack drone is... Join the Mission Alliance The utility for: ； In the formula: It is an attack drone The benefits of joining a mission alliance. This is the current utility of the Mission Alliance; S23: Defines the utility requirement of a mission coalition for a reconnaissance UAV, which is defined as the difference between the maximum utility of the mission coalition and the utility of the reconnaissance UAV before joining the mission coalition. For reconnaissance drones utility needs for: ； S24: Construct the individual cost calculation function for the attack drone, comprehensively considering distance and heading factors, and define the attack drone. Join the Mission Alliance Individual cost for: ； In the formula: It is the distance correlation coefficient. It is the velocity correlation coefficient. This represents the Euclidean distance between the attack drone and the mission target. It is the cosine angle between the velocity vector of the attack drone and the velocity vector of the mission target; S25: Construct the individual cost calculation function for reconnaissance UAVs, and define the reconnaissance UAV. Join the Mission Alliance Individual cost for: ； In the formula: This represents the Euclidean distance between the reconnaissance drone and the mission target. It is the optimal distance that reconnaissance drones and mission targets are expected to maintain. It is the cosine angle between the velocity vector of the reconnaissance UAV and the velocity vector of the mission target.

4. The heterogeneous UAV task allocation method based on task alliances and individual utility according to claim 3, characterized in that, The specific steps for step S3 are as follows: S31: Definition and initialization of attack drone state variables: Define the satisfaction state variable `satisfied` for the attack drone, which is a binary variable used to indicate whether the attack drone has found a satisfactory expected task; define the iteration count `Iter`, which is an integer variable used to compare the iteration counts of the attack drones to resolve possible selection conflicts during algorithm execution; define the self-alliance cognition `CS`, which is a set used to store the current task assignment status of all attack drones from the perspective of this attack drone; Before the algorithm starts, the above variables are initialized: satisfied is set to false, indicating that the currently selected task is not yet satisfactory; Iter is set to 0; CS is initialized to S0, that is, by default all attack drones select an empty task at the initial moment. S32: Determine the satisfaction state variable of the attack drone. If it is true, maintain the current state; if it is false, proceed to the mission selection phase. This phase involves iterating through all missions and calculating the utility of joining each mission coalition. for: ； All drones that choose an empty mission form an empty mission alliance. The utility of joining the empty mission alliance is defined as fixed at 0. Then, the drone with the highest utility is selected to join its mission alliance. At the same time, the iteration count Iter is incremented by 1, the satisfaction state variable satisfied is updated to true, and the alliance cognition CS is updated according to its own mission selection scheme. S33: The attack drone broadcasts its own alliance cognition CS and iteration number Iter to all neighboring attack drones within the communication topology. At the same time, it receives the alliance cognition and iteration number information broadcast by neighboring attack drones, and combines the received information with its own information to form a list of "iteration number - alliance cognition" information pairs to be processed. S34: Traverse the "iteration count - alliance cognition" information pair list obtained in step S33, and update the decision according to the following rules: Prioritize adopting alliance cognitive information with a larger iteration count (Iter); When the number of iterations is the same, compare the costs of the attack drone in the corresponding mission coalition. Prioritize adopting alliance cognitive information that has lower costs; If the neighbor's information is adopted according to the above rules, that is, if the neighbor's information is found to be better than its own information, then the neighbor's information is used to update its own "iteration count - alliance cognition" information to modify its own alliance cognition and iteration count, and its own satisfaction state variable satisfied is reset to false to trigger the next round of task selection; otherwise, no modification is made. S35: Repeat steps S32 to S34 until the satisfaction status variable satisfied of all attack drones in the cluster is true, which indicates that the task allocation at the attack drone level is complete. S36: After the attack drone completes the mission assignment, it will broadcast the final mission coalition to the reconnaissance drone.

5. The heterogeneous UAV task allocation method based on task alliances and individual utility according to claim 4, characterized in that, The specific steps for step S4 are as follows: S41: Based on the task alliance generation results in step S36 and the calculation formula in step S23, calculate the utility requirement of each task alliance for the reconnaissance UAV. And combine all utility demands into a task alliance bid list. ,set up Indicates joining the mission alliance For reconnaissance drones Utility needs; S42: Calculate the reconnaissance UAV's entry into the mission coalition based on the formula in step S25. The cost And all the cost combinations are defined as the cost list for reconnaissance drones. ,set up Reconnaissance drone Join the Mission Alliance The cost; S43: Create a task assignment list and the list of maximum utility ,set up Reconnaissance drone Choose to join the mission alliance , This indicates the utility of a reconnaissance UAV in selecting this mission coalition; initialize each value in both lists to 0. S44: Traverse the task assignment list The task is to calculate the utility of selecting this task. : ； Finally, select the task with the highest utility as the current target task and update its task allocation list accordingly. and the list of maximum utility ; S45: Sends its own mission assignment list to all neighboring reconnaissance drones. and the list of maximum utility It also receives the task assignment list and maximum utility list of all neighboring reconnaissance drones; S46: Traverse the task assignment list of neighboring reconnaissance drones and check for task selection conflicts; If a neighboring reconnaissance drone is found to be more effective for the same task, the neighboring reconnaissance drone is determined to win, and the task selection is abandoned. The task allocation list is then updated based on the information from the neighboring reconnaissance drone. and the list of maximum utility To resolve the conflict; S47: Check if the task allocation list of the drone itself has changed after step S45. If it has changed, it means that the task has been taken over by other reconnaissance drones. Continue to execute step S44 to reselect the task. If the task allocation lists of all reconnaissance drones have not changed, the task allocation of the heterogeneous drone cluster is completed.

Images