Communication-free aircraft cluster interception / collision task allocation method

Abstract

The invention provides a non-communication aircraft cluster interception / collision task allocation method, and aims to solve the problems of efficient task allocation and target interception of an aircraft cluster in a communication limited environment. Different from a traditional single target interception method, the multi-target interception method expands multi-target interception to a more complex task allocation scene, and significantly improves the interception success rate of an aircraft cluster facing a plurality of intrusion targets. Task allocation is performed by introducing a graph game framework, cooperative task allocation between aircrafts can be realized through local information under the condition of no communication, and the decision of each aircraft is ensured to be the optimal response to a neighbor strategy. According to the method, the effectiveness of the proposed algorithm in the multi-target interception task is verified, and the feasibility of the algorithm in practical application is consolidated. According to the method, an important technical means is provided for coping with multi-target intrusion threats, and development of an aircraft cluster autonomous task allocation and collaborative interception technology is promoted.

Description

Technical Field

[0001] This invention relates to key technologies for autonomous interception / collision of aircraft, including multi-aircraft task allocation, graph game modeling, line-of-sight perception, correlation equilibrium calculation, autonomous interception, and autonomous collision. Specifically, it introduces a graph game-based task modeling method and a communication-free correlation equilibrium solution method to address the task allocation and target coverage issues faced by aircraft swarms performing interception missions without direct communication. This encompasses technical areas such as graph game theory, task allocation, collaborative control, autonomous interception, and autonomous collision, aiming to enhance the multi-target interception capability and mission completion efficiency of drone swarms in communication-restricted environments. Background Art

[0002] With the widespread adoption of miniaturized and intelligent aircraft, drone swarms are playing an increasingly important role in both military and civilian applications. However, achieving efficient and robust task allocation within potentially hostile drone swarms in environments with limited or even no communication has become a key challenge in aerospace defense systems. Traditional interception methods, such as physical capture or frequency jamming, are ineffective against large, coordinated target swarms. Existing multi-robot task allocation algorithms, such as the Hungarian algorithm and the consensus-based auction algorithm (CBBA), generally rely on stable communication within the swarm to achieve information consistency and task coordination.

[0003] However, in real combat environments, enemies often use tactics like obstruction, evasive maneuvers, and electromagnetic interference to disrupt communication links between drones, putting communication-reliant swarm task allocation strategies at risk of failure. To address these challenges, some research has begun exploring distributed task allocation frameworks based on graph models. These frameworks leverage the local perception information of drones to construct adjacency graphs and incorporate the correlated equilibrium mechanism from game theory. This allows each drone to independently make optimal decisions in the absence of direct communication, thereby achieving task coordination.

[0004] However, existing methods still suffer from insufficient task coverage, slow allocation convergence, and weak local conflict resolution capabilities. Most methods also fail to fully integrate active observation strategies to enhance the effectiveness of local information. Furthermore, current game theory approaches focus on the existence and convergence of Nash equilibria, without systematically addressing the efficiency and feasibility of solving related equilibria under local observation constraints. Therefore, there is an urgent need to propose a method for allocating interception tasks for drone swarms that combines local feasibility, task coverage, rapid convergence, and adaptability to non-communication scenarios, in order to meet the needs of coordinated interception in complex environments. Summary of the Invention

[0005] The present invention proposes a method for allocating interception / collision tasks for a cluster of aircraft without communication, aiming to solve the problems of efficient task allocation and target interception for aircraft clusters in communication-restricted environments. Unlike traditional single-target interception methods, the present invention extends multi-target interception to more complex task allocation scenarios, significantly improving the interception success rate of aircraft clusters facing multiple intrusion targets. By introducing a graph game framework for task allocation, the present invention can achieve collaborative task allocation between aircraft through local information under non-communication conditions, ensuring that the decision of each aircraft is the optimal response to the neighbor's strategy. The present invention verifies the effectiveness of the proposed algorithm in multi-target interception tasks and consolidates its feasibility in practical applications. This method provides an important technical means for dealing with multi-target intrusion threats and promotes the development of autonomous task allocation and collaborative interception technology for aircraft clusters.

[0006] This paper first introduces the related interception / collision cluster and enemy cluster models, the worker and task list model, the graph game model, and the reward function model. Based on these models, this paper proposes a new graph game framework and constructs a correlated equilibrium (CE) within this framework. By solving the graph game problem, this paper successfully solves the task allocation problem for aircraft swarms and proposes an optimized task allocation scheme.

[0007] Intercept / Collision Cluster and Enemy Cluster: Let and Denote the set of aircraft in the interception / collision group and the set of intruders in the enemy group, respectively. The sizes of the interception / collision group and the enemy group are I and H, respectively. Considering I≠H, when the size of the interception / collision group does not match the size of the enemy group, it can be called a multi-tasking allocation problem, which can be divided into the following two cases:

[0008] When I < H, our interceptor / collision aircraft is equipped with reusable tools such as net guns. One interceptor / collision aircraft can be assigned to multiple invaders.

[0009] When I>H, multiple interception / collision aircraft are assigned to the same intruder to improve the interception / collision success rate and make full use of resources.

[0010] Worker and task lists: To solve the problem of mismatch between the interception / collision cluster and the enemy cluster, we introduce virtual workers and virtual tasks to construct worker lists and task lists. and Represents the worker list set and task list set respectively. The size of the worker list and task list are both n, where

[0011] Undirected graph model: is an undirected graph representing the observation structure of the worker list, such as Figure 1 As shown, is a list of workers, forming the endpoints of an undirected graph; is an edge set such that the jth worker has an edge e(i,j)∈ε if and only if the jth worker is within the observation range of the ith worker.

[0012] Worker List The CCP consists of n workers, and the i-th worker is defined as WL i , for each worker WL i There is a finite set of pure strategies Worker WL i Pure Strategy The global cooperation strategy of all workers is in, is the global strategy space.

[0013] Worker WL i The neighbor set of Among them, x i , x j Representing workers WL i and WL j The neighbor set includes all other workers in the panoramic camera range R and workers WL i Myself. The number of workers in is n i . Worker WL i The set of interceptable targets is Includes all tasks within the panoramic camera range R and the forward camera observation range C. Further, we have Represents worker WL i and worker WL j The intersection of neighbor sets, using a ij express Joint action of workers in the

[0014] Profit function: Worker WL i Neighborhood set The strategy is Indicates that the global strategy is in the neighbor set The projection on the neighbor set Remove worker WL i The strategy is in, Represents the strategy projection s i The local profit matrix is ​​composed of the profit u i (a,s -i ) It includes pure strategies and mixed strategies.

[0015] For the pure strategy, considering that the distance affects the urgency of the task, we have the profit function:

[0016]

[0017] The above formula shows that in the worker WL i Select task a, other workers in the neighbor set select strategy s -i The profit when . Where D is the upper bound of the distance, ensuring that the profit is positive; It is a worker WL i The distance from task a; is the conflict detection indicator function, when the strategy s i No conflict, If there is a conflict

[0018] For mixed strategies, we have the global strategy space The joint probability distribution p(s) on satisfies:

[0019]

[0020]

[0021] Worker WL i The local probability distribution is:

[0022]

[0023] Among them, p i (a,s -i ) represents the selection of s in the neighbor set -i Strategy, worker WL i The probability of selecting task a, Represents those that satisfy the neighbor set selection s -i The global policy of the policy s.

[0024] Therefore, the worker WL i In the probability set The distribution benefits are:

[0025]

[0026] Graph game model: The graph game model is in, is a set of n local payoff matrices, and G is the aforementioned undirected graph. Graph games can be formulated as an optimization problem:

[0027]

[0028] Among them, u i(P i ) for workers WL i Profit function.

[0029] For workers WL i and its neighbor set There is a local graph game model

[0030] The present invention proposes a method for allocating interception / collision tasks of a cluster of non-communication aircraft, and its planning flow chart is as follows: Figure 2 As shown, based on the previous definition, the implementation steps are as follows:

[0031] Step 1: Build a worker list and task list based on the interception / collision cluster and the enemy cluster

[0032] The main purpose of this step is to standardize and unify the original interceptor aircraft cluster and invader cluster to meet the needs of building the subsequent graph game model. Since the number of our aircraft I and the number of enemy invaders H may not be equal in reality, directly assigning tasks will lead to an unbalanced model. Therefore, this step introduces the concepts of "workers" and "tasks" and maps the actual aircraft and invaders into a list of workers of the same size. and task list This is done to ensure that the number of participants (workers) in the subsequent game model and the number of targets (tasks) to be assigned are equal, usually If we have fewer friendly aircraft than the enemy (I < H), some or all friendly aircraft may need to take on multiple tasks (this is achieved by replicating themselves into multiple virtual workers, each equipped with a reusable tool such as a net gun). If we have more friendly aircraft than the enemy (I > H), multiple friendly aircraft may coordinate to deal with a single invader (by treating a single invader as multiple virtual tasks, or by assigning excess aircraft to the same invader to increase the success rate). This preprocessing step is the basis for subsequent effective task allocation.

[0033] In step 1, based on the set of aircraft in a given interception / collision group and the invaders of the enemy cluster Build a list of workers and task list

[0034] S11. Input interceptor cluster and invader clusters

[0035] S12. Determine the size relationship between I and H. If I=H, jump to step S13. If I<H, jump to step S14. If I>H, jump to step S15.

[0036] S13, order task list Worker List

[0037] S14, order in is a floor function, then

[0038]

[0039] S15, order Then there is

[0040]

[0041] Step 2: Task allocation based on the distributed algorithm of correlation balance

[0042] This step aims to assign appropriate interception tasks to each worker without direct communication. The core idea is to leverage graph game theory and the concept of correlated equilibrium. Graph games allow us to decompose the complex global task allocation problem into a series of local decision-making problems, where each worker only needs to consider the behavior of its neighbors. Correlated equilibrium provides a mechanism that allows workers to coordinate their actions through a shared probability distribution, even without direct communication, to achieve a collectively optimal or near-optimal task allocation. This step will detail how to construct local games, define correlated equilibrium constraints, extend to the global game, and ultimately solve the task allocation solution using a distributed algorithm.

[0043] S21. Construct the relevant equilibrium of local games.

[0044] To achieve distributed decision making, we first focus on a single worker WL i and its neighbor set The task allocation problem in this local area can be modeled as a local graph game. Correlated equilibrium is the key solution concept of this local game. It describes a state in which there exists a joint probability distribution p i (s i )(s i (where is the joint strategy of worker i and its neighbors). Under the guidance of this distribution, any worker j (including worker i itself) who unilaterally deviates from their recommended strategy (action a) will not obtain a higher expected return than if they followed the recommended strategy. This property ensures that decisions based on this probability distribution are stable within a local scope, and no worker has the motivation to unilaterally change their behavior. Formula (11) is the core constraint of the correlated equilibrium.

[0045] S211. In a local graph game, the relevant equilibrium is given by the joint distribution p i (s i) indicates that, Correlated equilibrium for any And any strategy pair There are the following constraints:

[0046]

[0047] in, represents the jth worker choosing action a when u j expectations, represents the benefit u of the jth worker when he chooses action a′ j The expectation under the probability distribution of the actions taken by other workers based on the assumption that the jth worker chose action a.

[0048] S212. Calculate the expected local returns:

[0049]

[0050] S213. The local graph game model is represented as:

[0051]

[0052] To avoid the exponential growth of the strategy set, we cannot directly extend the local correlated equilibrium to the global game, so we will introduce consistency constraints below.

[0053] S22. Construct a global graph game model.

[0054] Although the local correlated equilibrium provides each worker with a basis for decision-making within its neighborhood, in order to ensure the global consistency and effectiveness of the actions of the entire aircraft cluster, these local decision probabilities need to be integrated. This step aims to build a global graph game model. The key lies in dealing with the overlap between the local games of different workers (i.e., the cross-neighborhood set). ). It is necessary to ensure that for common neighbors in overlapping neighborhoods, the marginal probability distributions followed in different local games are consistent. This is achieved by introducing consistency constraints (such as Equation 14). The goal of the global graph game is to find a global joint probability distribution p(s) (where s is the joint strategy of all workers) that not only satisfies the local correlated equilibrium conditions of all workers (such as Equation 11 or its equivalent form Equation 15), but also satisfies these consistency constraints, and usually aims to maximize some global performance indicator, such as the sum of the expected returns of all workers (as shown in Equation 16).

[0055] S221. To ensure consistency between decisions of cross-neighborhood sets, define consistency constraints:

[0056]

[0057] S222. Equation (11) is equivalently expressed as:

[0058]

[0059] S223. The global graph game is represented as:

[0060]

[0061] S23. Solving graph games based on distributed algorithms.

[0062] Directly solving the aforementioned global graph game model (Formula 16 and its constraints) may involve extremely high computational complexity, especially in large-scale clusters. Therefore, this step proposes a distributed, iterative solution algorithm to approximate or find the solution to the graph game. The core idea of ​​the algorithm is to process the worker list one by one in a predetermined order (for example, according to the topological position of the workers in the graph G, such as the distance from the center of the graph). Each worker in WL i When processing a worker WL i When it solves a local graph game problem (as shown in Formula 13), the problem aims to maximize its local expected benefits while satisfying the relevant equilibrium constraints. Importantly, if some of the worker's neighbors have been processed, then when solving its local game, the decisions of these processed neighbors (determined probability distributions or specific actions) will be used as known conditions or constraints (such as the conditional probability calculation method shown in Formula 18) to ensure the sequential consistency of the decision. Through this sequential processing method, the global probability set of the entire cluster is gradually constructed.

[0063] S231. Input the worker's graph structure and local payoff matrix.

[0064] S232, according to the distance from the center of the graph G Sort the elements in .

[0065] S233, for each element WL in the work list i , solving formula (13) yields:

[0066]

[0067] S2331, determine WL i Neighbor Set Are all the elements in unprocessed? If so, jump to step S2332; if not, jump to step S2333.

[0068] S2332, let the global probability set

[0069] S2333, set neighbor set The processed collection is For collection The elements in the selected action are:

[0070]

[0071] make parameter Indicates that in the collection The element in has selected action The joint probability of the case.

[0072] The advantages and benefits of this invention include: It proposes a method for allocating interception / collision tasks for a swarm of non-communication aircraft, solving the task allocation problem in communication-restricted environments. By leveraging a graph game framework and solving related equilibrium problems, it optimizes the task allocation of swarm aircraft in multi-target interception, improves the interception success rate, and promotes the development of autonomous, collaborative interception technology for swarm aircraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a schematic diagram of the coordinate system and camera imaging model.

[0074] Figure 2 It is a flow chart of the method for allocating interception / collision tasks for a cluster of non-communication aircraft. DETAILED DESCRIPTION

[0075] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0076] Example: The existing interceptor cluster consists of 2 aircraft and the enemy cluster consists of 3 intruders. Assume that their initial positions (unit: meters) in a two-dimensional plane are given as follows:

[0077] Interceptor 1: (2,8); Interceptor 2: (7,7);

[0078] Invader 1: (3,3); Invader 2: (8,4); Invader 3: (5,10);

[0079] Assume that the sensor's observation range parameters are: neighbor detection range R = 6 meters, target detection range C∪R (simplified as radius) = 7 meters, and the benefit function parameter D = 15.

[0080] The implementation steps of the method for allocating interception tasks of a cluster of non-communication aircraft of the present invention are as follows:

[0081] Step 1: Build a worker list and task list based on the interception / collision cluster and the enemy cluster

[0082] S11. Input interceptor cluster and invader clusters location information.

[0083] S12. Determine the magnitude relationship between I and H. Since I=2 and H=3, I<H, and the process goes to step S14.

[0084] S14. Calculation parameters According to formula (8), the theoretical worker list Should be I × 2 = 4 workers. Task list According to formula (7), Contains 3 tasks. Due to the requirement of workers list and task list The scale of n = max(I,H) = 3 is equal, so the final list of workers constructed is Contains 3 workers (discarding the last virtual interceptor), task list It contains 3 tasks. The details are as follows:

[0085] Worker List

[0086] Worker 1: ID = 1, original interceptor is interceptor 1, position is (2, 8);

[0087] Worker 2: ID = 2, the original interceptor is interceptor 1, the position is (2, 8), and the virtual interceptor is copied from interceptor 1;

[0088] Worker 3: ID = 3, original interceptor is interceptor 2, position is (7,7);

[0089] Task List

[0090] Task 1: ID = 1, the original intruder is Intruder 1, and the location is (3, 3);

[0091] Task 2: ID = 2, the original invader is invader 2, and the location is (8, 4);

[0092] Task 3: ID = 3, the original invader is invader 3, and the location is (5, 10);

[0093] Step 2: Task allocation based on the distributed algorithm of correlation balance

[0094] S21. Construct the relevant equilibrium of the local game. First, determine each worker's neighbor set (NWL) and attackable target set (NTL).

[0095] Calculate the distance between workers and get the neighbor set

[0096]

[0097] Calculate the distance from the worker to the task and determine the set of attackable targets

[0098]

[0099] Calculating the profit function Among them, D = 15, if the joint action s i There is no conflict (ie W0, W1, W2 choose different tasks), then otherwise

[0100] The relevant equilibrium constraints of the local game are defined according to formula (11).

[0101] S22. Construct a global graph game model and introduce a consistency constraint (Formula (14)). When processing workers sequentially, the post-processing workers need to ensure that the probability distribution they calculate is consistent with the marginal probability of their processed neighbors on their intersection.

[0102] S23. Graph game solving based on distributed algorithms.

[0103] S231. Input worker list Task List Neighbor Relations Attackable targets and the profit function parameter D.

[0104] S232. Process in order of distance from the center of the distance graph: worker 1, worker 2, worker 3.

[0105] S233. Solve the local graph game (Formula (17)) for each worker and obtain its probability distribution p i .

[0106] Processing Worker 1:

[0107] Neighbor Set

[0108] S2331: Determine the neighbor processing status: Worker 2 and Worker 3 have not processed it.

[0109] S2332: Jump here. Solve the local game LP problem to maximize the expected payoff of worker 1, subject to the CE conditions of all neighbors (Formula (11)).

[0110] Solve to obtain the probability distribution of worker 1's joint strategy p1(s1). The optimal (or near-optimal) joint strategy and its probability are calculated as: (where s1 = (worker 1 action, worker 2 action, worker 3 action));

[0111] p1((3,1,2))=1(other probabilities are extremely small);

[0112] Store p1 in the global probability set middle.

[0113] Processing Worker 2:

[0114] Neighbor Set

[0115] S2331: Determine the neighbor processing status: Worker 1 has processed it.

[0116] S2333: Jump here. Processed neighbor set Solve the local game LP problem to maximize the expected payoff of worker 2, subject to the CE condition of all neighbors (Formula (11)) and the consistency constraint with worker 1 (Formula (14)). The consistency constraint requires that the marginal probability of p2(s2) on the common neighbors {1,2,3} is consistent with p1(s1).

[0117] Solve to obtain the probability distribution of worker 2's joint strategy The calculated optimal joint strategy and its probability are: p2((3,1,2))=1 (other probabilities are extremely small);

[0118] Store p2 in the global probability set middle.

[0119] Processing Worker 3:

[0120] Neighbor Set

[0121] S2331: Determine the neighbor processing status: Worker 1 and Worker 2 have processed.

[0122] S2333: Jump here. Processed neighbor set Solve the local game LP problem and maximize the expected reward of worker 3, with the constraints of CE conditions of all neighbors (Formula (11)) and consistency constraints with workers 1 and 2 (Formula (14)).

[0123] Solve to obtain the probability distribution of worker 3's joint strategy The calculated optimal joint strategy and its probability are: p3((3,1,2))=1 (other probabilities are extremely small);

[0124] Store p3 into the global probability set middle.

[0125] So far, by executing the solution steps for all workers (worker 1, worker 2, worker 3) in the embodiment, we have constructed a complete global probability set This collection represents a collaborative, conflict-free probabilistic solution for the entire interceptor cluster on how to allocate tasks in the current interception scenario. During the actual flight and interception process, each original interceptor (and its corresponding virtual worker) will The probability distribution defined in is used to independently and randomly select the intruder (task) to attack. It is based on related equilibrium calculations. This distributed decision-making method can guide the cluster to achieve the optimization of the overall mission objectives without direct communication, effectively improve the success rate and efficiency of multi-target interception, and adapt to complex environments with limited or even no communication.

Claims

1. A method for allocating interception / collision tasks for a cluster of non-communication aircraft, characterized in that: Here are the steps: Step 1: Build a worker list and task list based on the interception / collision cluster and the enemy cluster The original interceptor aircraft cluster and invader cluster are standardized and unified to meet the needs of building the subsequent graph game model. Since the number of our aircraft I and the number of enemy invaders H may not be equal, directly assigning tasks will lead to an unbalanced model. Therefore, the concepts of "workers" and "tasks" are introduced to map the actual aircraft and invaders into a list of workers with the same size. and task list Ensure that the number of participants in the game model, i.e., the number of workers, and the number of tasks to be assigned, i.e., the number of tasks to be assigned, are equal. If our aircraft are fewer than the enemy's, that is, I < H, then some or all of our aircraft may need to undertake multiple missions; if our aircraft are more than the enemy's, that is, I > H, then multiple of our aircraft may cooperate to deal with one invader; Step 2: Task allocation based on the distributed algorithm of correlation balance Introducing graph game theory and the concept of correlated equilibrium; In the absence of direct communication, each worker is assigned an appropriate interception task; graph games allow the complex global task allocation problem to be decomposed into a series of local decision-making problems, and each worker only needs to consider the behavior of its neighbors; correlated equilibrium allows workers to coordinate their actions through a shared probability distribution even without direct communication, achieving optimal or near-optimal task allocation results.

2. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 1, characterized in that: In step 1, based on the set of aircraft in a given interception / collision group and the invaders of the enemy cluster Build a list of workers and task list S11. Input interceptor cluster and invader clusters S12, judging the relationship between I and H, if I=H, jumping to step S13, if I<H, jumping to step S14, if I>H, jumping to step S15; S13, order task list Worker List S14, order in is a floor function, then S15, order Then there is 3. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 1 or 2, characterized in that: In step 2, the method further includes the following steps: S21. Constructing the relevant equilibrium of local games; To achieve distributed decision making, focus on a single worker WL i and its neighbor set The task allocation problem in this local area is modeled as a local graph game; correlated equilibrium is the key solution concept of this local game; there is a joint probability distribution p i (s i ), s i is the joint strategy of worker i and its neighbors. Any worker j who unilaterally deviates from the recommended strategy will not obtain a higher expected return than following the recommended strategy. S22. Construct a global graph game model; Construct a global graph game model; the key is to deal with the overlapping parts between different workers' local games, namely the cross-neighborhood set It is necessary to ensure that for common neighbors in overlapping neighborhoods, the marginal probability distributions followed by them in different local games are consistent; this is achieved by introducing consistency constraints; S23. Solving graph games based on distributed algorithms; Introduce a distributed, iterative solution algorithm to approximate or find the solution of the graph game; process the worker list one by one Each worker in WL i ; When processing a worker WL i When a worker is in a certain situation, it will solve a local graph game problem to maximize its local expected benefits while satisfying the relevant equilibrium constraints. If some of the worker's neighbors have been processed, then when solving its local game, the decisions of these processed neighbors will be used as known conditions or constraints to ensure the sequential consistency of the decision. Through this sequential processing method, the global probability set of the entire cluster is gradually constructed.

4. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 3, characterized in that: In step S21, the worker WL i Neighborhood set The strategy is Indicates that the global strategy is in the neighbor set The projection on the neighbor set Remove worker WL i The strategy is in, Represents the strategy projection s i The local profit matrix is ​​composed of the profit u i (a,s -i ) composition; including pure strategies and mixed strategies.

5. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 4, characterized in that: For pure strategies, considering that distance affects the urgency of the task, there is a profit function: The above formula shows that in the worker WL i Select task a, other workers in the neighbor set select strategy s -i The profit when ; where D is the upper bound of the distance, ensuring that the profit is positive; It is a worker WL i The distance from task a; is the conflict detection indicator function, when the strategy s i No conflict, If there is a conflict 6. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 4, characterized in that: For mixed strategies, there is a global strategy space The joint probability distribution p(s) on satisfies: Worker WL i The local probability distribution is: Among them, p i (a,s -i ) represents the selection of s in the neighbor set -i Strategy, worker WL i The probability of selecting task a, Represents those that satisfy the neighbor set selection s -i The global strategy of the strategy s; Therefore, the worker WL i In the probability set The distribution benefits are:

7. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 6, characterized in that: In step S21, specifically: S211. In a local graph game, the relevant equilibrium is given by the joint distribution p i (s i ) indicates that, Correlated equilibrium for any And any strategy pair There are the following constraints: in, represents the jth worker choosing action a when u j expectations, represents the benefit u of the jth worker when he chooses action a′ j The expectation under the probability distribution of the actions taken by other workers based on the assumption that the jth worker chose action a; S212. Calculate the expected local returns: S213. The local graph game model is represented as:

8. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 7, characterized in that: In step S22, specifically: S221. To ensure consistency between decisions of cross-neighborhood sets, define consistency constraints: S222. Equation (11) is equivalently expressed as: S223. The global graph game is represented as:

9. The method for allocating interception / collision tasks for a swarm of non-communication aircraft according to claim 8, characterized in that: In step S23, specifically: S231. Input the worker's graph structure and local payoff matrix; S232, according to the distance from the center of the graph G Sort the elements in ; S233, for each element WL in the work list i , solving formula (13) yields: S2331, determine WL i Neighbor Set Are all elements in unprocessed? If so, jump to step S2332; if not, jump to step S2333; S2332, let the global probability set S2333, set neighbor set The processed collection is For collection The elements in the selected action are: make parameter Indicates that in the collection The element in has selected action The joint probability of the case.

Images