Defensive allocation methods, devices, and electronic equipment for multi-agent adversarial games

Abstract

This application provides a defense allocation method, apparatus, and electronic device for multi-agent adversarial games. The method includes selecting multiple candidate defense alliances from the unassigned cluster of defender agents at the current time; based on a constructed zero-sum game objective function, the attacker aims to maximize the number of attacker agents invading the target protected area, while the defender aims to minimize the number of attacker agents successfully invading the target protected area. The method iteratively searches for a target defense alliance among the candidate alliances and generates an initial defense allocation strategy based on the target defense alliance and the attacker agents it defends against; using monotonic constraints, the initial defense allocation strategy is optimized to obtain the final defense allocation strategy; the monotonic constraint is that the number of attacker agents expected to be unable to reach the target protected area at the current time increases compared to the previous time. This invention balances the optimality and efficiency of each defense allocation while ensuring continuous optimization of defense performance.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a defensive allocation method, apparatus, and electronic device for multi-agent adversarial games. Background Technology

[0002] In typical multi-agent adversarial game scenarios, the attacking and defending sides engage in a battle around a target set under conditions of unknown strategy. Specifically, the attacking team maximizes the number of times the attacker can successfully infiltrate the target area through collaborative decision-making, while avoiding interception by the defender; the defending team, on the other hand, needs to build a dynamic defense system to minimize the number of attackers entering the target set.

[0003] Existing defensive allocation methods for multi-agent adversarial games include bipartite graph matching algorithms based on a single defender, traversal allocation algorithms, and heuristic hierarchical allocation algorithms. These methods have the following problems:

[0004] 1. Insufficient collaboration: The matching algorithm based on a single defender does not consider the collaboration among defenders, resulting in an overly simplistic cooperation model. Real-world situations are diverse and changeable, making it difficult to apply to real-world scenarios.

[0005] 2. High computational overhead: When the number of agents is large, the traversal allocation algorithm traverses all combinations of sub-alliances, which is computationally intensive and makes it difficult to deploy online.

[0006] 3. Insufficient robustness: It only performs static allocation based on the initial joint state of the attacking and defending teams, and cannot be updated in real time based on feedback, making it difficult to cope with real-time changes in the attacker's strategy and the uncertainty of the environment.

[0007] 4. Lack of continuous optimization mechanism: The heuristic hierarchical allocation algorithm can reduce the amount of computation to a certain extent, but the solution obtained by the heuristic method is suboptimal. It lacks continuous performance optimization during the allocation process and cannot guarantee the overall improvement of defense performance. Summary of the Invention

[0008] In view of this, the purpose of the present invention is to provide a defense allocation method, apparatus and electronic device for multi-agent adversarial games, which aims to ensure the continuous optimization of defense performance while taking into account the optimality and efficiency of each defense allocation.

[0009] Firstly, a defense allocation method for multi-agent adversarial games is provided, wherein the multi-agent group includes a cluster of defending agents and a cluster of attacking agents, the cluster of defending agents consisting of several defending agents; the cluster of attacking agents consisting of several attacking agents; the method includes:

[0010] Several defensive agents are selected from the unassigned defensive agent cluster at the current moment to form multiple candidate defense alliances; each candidate defense alliance includes several defensive agents.

[0011] Based on the constructed zero-sum game objective function, the attacker aims to maximize the number of attacking agents that infiltrate the target protected area, while the defender aims to minimize the number of attacking agents that successfully infiltrate the target protected area. The attacker iteratively searches for a target defense alliance among the candidate defense alliances and generates an initial defense allocation strategy based on the target defense alliance and the attacking agents it defends against. The defense allocation strategy is used to assign a defense alliance to each attacking agent.

[0012] By using monotonic constraints, the initial defense allocation strategy generated at the current moment is optimized to obtain the final defense allocation strategy; the monotonic constraint is that the number of attack agents that are expected to be unable to reach the target protection area at the current moment increases compared to the previous moment.

[0013] Optionally, multiple candidate defense coalitions can be selected from the currently unassigned cluster of defender agents, including:

[0014] Obtain the secure activity area of ​​the attacker's intelligent agent cluster and the set of boundary points of the secure activity area;

[0015] Select the location point closest to the target protected area from the set of boundary location points;

[0016] The defensive agents that simultaneously reach the nearest location with the attacking agent are identified as the proactive defense alliance;

[0017] Select candidate defense alliances from the proactive defense alliances.

[0018] Optionally, candidate defense alliances can be selected from the active defense alliances, including:

[0019] In an active defense alliance, an initial active defense alliance is selected, and the number of defensive agents in the initial active defense alliance is no greater than the dimension of its spatial domain.

[0020] In the initial active defense alliance, select the defense alliance with the fewest defenders and identify it as a candidate alliance.

[0021] Optionally, based on the constructed zero-sum game objective function, the attacker aims to maximize the number of attacking agents that infiltrate the target protected area, while the defender aims to minimize the number of attacking agents that successfully infiltrate the target protected area. The process involves searching for a target defense alliance among candidate defense alliances and generating an initial defense allocation strategy based on the target defense alliance and the attacking agents it defends against, including:

[0022] Based on the constructed zero-sum game objective function, iterative optimization is performed to find the solution;

[0023] In each iteration, the attacker aims to maximize the number of attack agents that infiltrate the target protected area, while the defender aims to minimize the number of attack agents that successfully infiltrate the target protected area. They search for a target defense alliance and generate a defense allocation strategy based on the target defense alliance and the attack agents it defends against.

[0024] The defense allocation matrix is ​​determined based on the defense allocation strategy obtained in each iteration;

[0025] The defense allocation matrices obtained from each iteration are superimposed, and the initial defense allocation strategy is determined based on the final superimposed defense allocation matrix.

[0026] Optionally, the objective function of the zero-sum game is:

[0027]

[0028] in, This indicates the number of attackers expected to be unable to reach the target protected area. Represents the defense allocation matrix. This represents the joint state of all attacking agents and all defending agents; Indicates the sequence number of the defense alliance; Indicates the number of defense coalitions; Indicates the sequence number of the attacking agent; Indicates the number of attacking agents; Indicates the number of defensive agents; Indicates the serial number of the defensive intelligent agent; Represents the elements of the correlation matrix.

[0029] Optionally, by utilizing monotonic constraints, the initial defense allocation strategy generated at the current time can be optimized to obtain the final defense allocation strategy, including:

[0030] Based on the initial defense allocation strategy at the current moment, predict the number of attacking agents that cannot reach the target protected area at the current moment;

[0031] Determine whether the number of attack agents that are predicted to be unable to reach the target protected area at the current moment is greater than the number of attack agents at the previous moment.

[0032] If it is greater than, then the initial defense allocation strategy at the current moment will be determined as the final defense allocation strategy at the current moment;

[0033] If it is less than or equal to, then the final defense allocation strategy of the previous moment will be determined as the final defense allocation strategy of the current moment.

[0034] Optionally, if there are still attacking agents that have not been assigned to a defense coalition, the method also includes:

[0035] Based on a greedy algorithm, each unassigned attacking agent is assigned the nearest defense alliance.

[0036] Secondly, a defensive allocation device for multi-agent adversarial games is provided, wherein the multi-agent group includes a cluster of defending agents and a cluster of attacking agents, the cluster of defending agents consisting of several defending agents; the cluster of attacking agents consisting of several attacking agents; the device includes:

[0037] The selection unit is used to select several defensive agents from the unassigned defensive agent cluster at the current moment to form multiple candidate defense alliances; each candidate defense alliance includes several defensive agents.

[0038] The search unit is used to iteratively search for a target defense alliance among candidate defense alliances based on a constructed zero-sum game objective function. The attacker aims to maximize the number of attacking agents that infiltrate the target protected area, while the defender aims to minimize the number of attacking agents that successfully infiltrate the target protected area. An initial defense allocation strategy is generated based on the target defense alliance and the attacking agents it defends against. The defense allocation strategy is used to assign a defense alliance to each attacking agent.

[0039] The optimization unit is used to optimize the initial defense allocation strategy generated at the current moment using monotonic constraints to obtain the final defense allocation strategy. The monotonic constraint is that the number of attack agents that are expected to be unable to reach the target protection area at the current moment increases compared to the previous moment.

[0040] Thirdly, an electronic device is provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0041] The memory is used to store computer programs;

[0042] When the processor executes a program stored in the memory, it implements any of the steps of the method described in the first aspect.

[0043] Fourthly, a computer-readable storage medium is provided, wherein a computer program is stored therein, and when executed by a processor, the computer program implements the steps of any of the methods described in the first aspect.

[0044] This invention provides a defense allocation method, apparatus, and electronic device for multi-agent adversarial games. The method involves selecting multiple candidate defense alliances from the current unassigned cluster of defensive agents. Each candidate alliance includes several defensive agents. Based on a constructed zero-sum game objective function, the attacker aims to maximize the number of attackers invading the target protected area, while the defender aims to minimize the number of attackers successfully invading the target protected area. The method iteratively searches for a target defense alliance among the candidate alliances and generates an initial defense allocation strategy based on the target defense alliance and the attackers it defends against. Using monotonic constraints, the initial defense allocation strategy generated at the current time is optimized to obtain the final defense allocation strategy. The monotonic constraint is that the number of attackers expected to be unable to reach the target protected area at the current time increases compared to the previous time. This invention reduces the size of feasible defense alliances by screening candidate alliances, significantly reduces the number of calculations and improves allocation efficiency by searching for the target defense alliance. Furthermore, the iterative search ensures the optimality of the defense allocation. The monotonic constraint method further optimizes the initial defense allocation strategy, ensuring that defense performance continuously improves over time.

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This diagram illustrates the overall framework of the defense allocation method provided in an embodiment of the present invention.

[0048] Figure 2 A flowchart of a defense allocation method for multi-agent adversarial games provided by an embodiment of the present invention is shown;

[0049] Figure 3 A schematic diagram of an active defense alliance according to an embodiment of the present invention is shown;

[0050] Figure 4 A flowchart illustrating a hierarchical integer linear programming method according to an embodiment of the present invention is shown.

[0051] Figure 5 A schematic diagram of the monotonic defense enhancement allocation process according to an embodiment of the present invention is shown;

[0052] Figure 6 This diagram illustrates the structure of a defensive allocation device for multi-agent adversarial games provided in an embodiment of the present invention.

[0053] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present invention is shown. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0055] This invention provides a defense allocation method for multi-agent adversarial games, applicable to real-world scenarios such as military confrontation and unmanned security. For example, in a military confrontation scenario, the agents are drones. If an enemy drone attempts to intrude into the defender's military target, the defender's drone fleet can employ a cooperative defense strategy to encircle and expel it, ensuring the safety of the military target. In an unmanned security scenario, multiple robots can work together to expel or encircle unauthorized personnel, preventing them from entering the protected area.

[0056] In multi-agent game adversarial problems, the attacking agent cluster usually aims to maximize the number of times it invades a specified target area, while the defending agent cluster needs to minimize the number of attackers who successfully invade the target, forming a typical zero-sum game.

[0057] To address the problems of insufficient cooperation, high computational overhead, insufficient robustness, and lack of continuous optimization mechanisms in current multi-agent game adversarial collaborative defense allocation, this invention provides a dynamic collaborative defense allocation algorithm that aims to balance the optimality and efficiency of each allocation while ensuring continuous optimization of overall defense performance.

[0058] This invention mainly includes two main steps, such as Figure 1The diagram illustrates the overall framework of the defense allocation method of this invention. The first step involves using a hierarchical integer linear programming approach to generate a defense allocation matrix, balancing optimality and efficiency. The second step, building upon the first, considers monotonicity constraints to ensure a steady improvement in defense performance. Finally, a heuristic greedy allocation method is employed to assign interception targets to the remaining defenders, generating the final defense allocation matrix.

[0059] The following is a detailed explanation using specific examples.

[0060] This invention provides a defense allocation method for multi-agent adversarial games. In the multi-agent adversarial game scenario, the multiple agents include a cluster of defending agents and a cluster of attacking agents. The defending agent cluster consists of several defending agents; the attacking agent cluster consists of several attacking agents. Figure 2 As shown, the method includes the following steps:

[0061] Step S201: Select several defense agents from the unassigned defense agent cluster at the current moment to form multiple candidate defense alliances.

[0062] Taking a military confrontation scenario as an example, since the drone swarms of the defending and attacking sides are dynamically changing, this application embodiment adopts a dynamic defense allocation strategy, which is periodically updated according to a preset time step. At each update moment, the defending agent swarms that have not yet been allocated, as well as the attacking agents that have not yet been allocated defensive agents, are first identified, and they are used as defense allocation objects to generate a defense allocation strategy.

[0063] In existing technologies, traversing all feasible defense sub-alliances to find the defense alliance-attack pair that achieves the optimal performance metrics is computationally intensive, wasting computing power and leading to inefficiency. Therefore, this invention employs a screening process within the defense agent cluster, selecting only those defenders capable of substantially preventing attackers from entering the target area before taking action. This strategy helps optimize resource utilization and improve defense efficiency.

[0064] In one feasible implementation, selecting multiple candidate defense alliances from the currently unassigned cluster of defender agents includes:

[0065] Step S201A: Obtain the secure activity area of ​​the attacker's intelligent agent cluster and the set of boundary points of the secure activity area.

[0066] In this step, the safe activity area is the range that the attacking agent can reach without being intercepted by the defending agent.

[0067] In one example, the scope of an attacking agent's secure activity area can be represented as a securely reachable set. In drone attack and defense scenarios, security can reach a set To attack drones Definitely not vulnerable to defensive drones The scope of the captured safe activity area.

[0068] Step S201B: Select the location point closest to the target protected area from the set of boundary location points.

[0069] In this embodiment of the invention, the location point is a convex point on the boundary of the safe reachable set, and this point is selected as the waypoint. The purpose is to maximize the effectiveness of the defense while minimizing the time or resources required for the defender to reach that location. This waypoint effectively prevents attackers from approaching the target's protected area.

[0070] Step S201C: The defensive agents that simultaneously reach the nearest position point with the attacking agent are identified as the active defense alliance.

[0071] Successful interception is determined when the defending agent and the attacking agent arrive at the closest point to the target's protected area at the same time; that is, they arrive simultaneously. Each proactive defense alliance corresponds to one attacking agent.

[0072] In one example, such as Figure 3 As shown, assuming the defending agent and the attacking agent have the same speed, the attacker... Arrive at the location The distance, and the defender , The capture circle perimeter to the position The distances are equal, meaning the minimum time required for them to reach that position at maximum speed is also the same; therefore, the defenders , In position Ensures protection against attackers If the interception is successful, then the attacker will be targeted. The proactive defense alliance is .

[0073] In a drone attack and defense scenario, multiple defensive drones intercept a single attacking drone. Some of these defensive drones can be positioned closest to the target drone within the attacking drone's safe operating range. These defensive drones, which ensured successful interception, played a substantial defensive role and formed an active defense alliance.

[0074] Step S201D: Select a candidate defense alliance from the active defense alliances.

[0075] In one feasible implementation, selecting candidate defense alliances from proactive defense alliances includes:

[0076] In an active defense alliance, an initial active defense alliance is selected, and the number of defensive agents in the initial active defense alliance is no greater than the dimension of its spatial domain.

[0077] In the initial active defense alliance, select the defense alliance with the fewest defenders and identify it as a candidate alliance.

[0078] For example, in a three-dimensional space with a dimension of 3, an initial active defense alliance is selected with no more than 3 defensive agents. From this initial alliance, the alliance with the fewest defenders is then chosen as a candidate alliance. This is because in three-dimensional space, allocating one defensive agent per dimension, or at most three, achieves comprehensive defense protection; too many agents would be a waste of resources. By selecting candidate alliances in this way, defense allocation is optimized.

[0079] Step S202: Based on the constructed zero-sum game objective function, the attacker aims to maximize the number of attack agents that invade the target protected area, while the defender aims to minimize the number of attack agents that successfully invade the target protected area. The attacker iteratively searches for the target defense alliance among the candidate defense alliances and generates an initial defense allocation strategy based on the target defense alliance and the attack agents it defends.

[0080] The defense allocation strategy is used to assign a defense alliance to each attacking agent.

[0081] In one feasible implementation, step S202 specifically includes the following steps:

[0082] Step S202A: Based on the constructed zero-sum game objective function, perform iterative optimization to find the solution.

[0083] In this embodiment of the invention, the zero-sum game objective function is as follows:

[0084] (1)

[0085] in, This indicates the number of attackers expected to be unable to reach the target protected area. Represents the defense allocation matrix. This represents the joint state of all attacking agents and all defending agents; Indicates the sequence number of the defense alliance; Indicates the number of defense coalitions; Indicates the sequence number of the attacking agent; Indicates the number of attacking agents; Indicates the number of defensive agents; Indicates the serial number of the defensive intelligent agent; Represents the elements of the correlation matrix.

[0086] By applying the constraints described above, each element i in the defense allocation matrix is ​​restricted to a binary variable of 0-1.

[0087] Step S202B: In each iteration, the attacker aims to maximize the number of attack agents that infiltrate the target protected area, while the defender aims to minimize the number of attack agents that successfully infiltrate the target protected area. They iteratively search for a target defense alliance and generate a defense allocation strategy based on the target defense alliance and the attack agents it defends against.

[0088] Step S202C: Determine the defense allocation matrix based on the defense allocation strategy obtained in each iteration.

[0089] The target defense coalition and its corresponding attacking agent are combined to form a defense coalition-attack pair. ( ) ,in, For the defense alliance, For attacking agents; multiple attack pairs form a defense allocation matrix. .

[0090] Step S202D: Superimpose the defense allocation matrices obtained in each iteration, and determine the initial defense allocation strategy based on the final superimposed defense allocation matrix.

[0091] In one example, the suboptimal solution of each iteration is assigned to a matrix. This indicates that, and will be in each iteration By superimposing these values, we obtain the initial defense allocation matrix. This initial defense allocation matrix corresponds to the initial defense allocation strategy.

[0092] The steps S201-S202 described above constitute the hierarchical integer linear programming method mentioned earlier. Hierarchical design is achieved by selecting candidate defense coalitions from all defensive agent clusters and then selecting the target defense coalition from among the candidate defense coalitions. The optimal solution for the linear programming problem is then found within the candidate defense coalitions.

[0093] Specifically, such as Figure 4 The diagram illustrates the specific process of hierarchical integer linear programming, which is as follows:

[0094] Step 1: Initialization: Let the initial defense allocation matrix be... Unassigned defense alliances are The set of attack agents that can be defended against but have not been assigned is In the initial defense allocation matrix, each attacking agent... Corresponding to a defense alliance Form a defensive alliance - attack against .

[0095] Step 2: Judgment and Check if all are not empty sets. If so, proceed to step three; otherwise, end the process.

[0096] Step 3: Traversal All members To determine whether the single-attack coordinated defense mission is expected to succeed, i.e. Is it feasible? If not, then... Remove This indicates an attack on the intelligent agent. Interception failed; if feasible, proceed to step four.

[0097] In this step, if If the attacker's security activity range does not overlap with the target protected area, then... It can successfully complete a single attack-defense coordinated mission, known as It is feasible.

[0098] Step 4: Take Active defense set Take active defense set irreducible subsets in .

[0099] In this step, the irreducible subset This is the candidate defense alliance.

[0100] Step 5: In Find the optimal solution for the linear programming problem.

[0101] Step 6: Superimpose the optimal solution onto the initial defense allocation matrix. middle.

[0102] Step 7: Already allocated and Remove them separately and , for The members of the group.

[0103] In existing technologies, it is necessary to traverse all possible combinations of defensive alliances, which requires a significant amount of computation. In the worst case, the number of calculations is... Where N is the number of defensive agents. The hierarchical integer linear programming method used in this application actively solves for the active and candidate defense alliances of the defense alliance through a heuristic approach, and then performs linear programming on this basis, significantly improving computational efficiency. The maximum number of computations is... , where n is the dimension of the spatial domain. This represents the number of active attackers. Furthermore, iterative calculations ensure a certain degree of optimality in the results.

[0104] Step S203: Optimize the initial defense allocation strategy generated at the current moment using the monotonic constraint to obtain the final defense allocation strategy; the monotonic constraint is that the number of attack agents that are expected to be unable to reach the target protection area at the current moment increases compared to the previous moment.

[0105] In one feasible implementation, the final defense allocation strategy is obtained by optimizing the initial defense allocation strategy generated at the current time using monotonic constraints, including:

[0106] Step S203A: Based on the initial defense allocation strategy at the current moment, predict the number of attacking agents that cannot reach the target protected area at the current moment.

[0107] Based on the above steps, by applying monotonicity constraints, it can be ensured that the number of attackers who fail to reach the target protection area increases over time, that is, the number of attackers successfully intercepted increases, thereby ensuring the continuous optimization of the defense allocation strategy of this application embodiment over time.

[0108] Step S203B: Determine whether the number of attack agents that are predicted to be unable to reach the target protected area at the current moment is greater than the number of attack agents at the previous moment.

[0109] Step S203C: If it is greater than, then the initial defense allocation strategy at the current moment is determined as the final defense allocation strategy at the current moment.

[0110] Step S203D: If it is less than or equal to, then the final defense allocation strategy of the previous moment is determined as the final defense allocation strategy of the current moment.

[0111] In a specific example, such as Figure 5 As shown, the initialized defense allocation matrix is Active attackers are The defense allocation matrix calculated using hierarchical integer linear programming is: The defense allocation matrix of the previous moment .

[0112] A hierarchical integer linear programming calculation is performed at each time step. If the number of attackers who are expected to be unable to reach the target set increases compared to the previous step, the result of the hierarchical integer linear programming is used. Update the initial defense allocation matrix Otherwise, retain the allocation matrix from the previous time step. ,Right now .

[0113] Furthermore, the number of active attackers may change between two consecutive time points. If the number of attackers who are expected to be unable to reach the target set has not increased compared to the previous step, it is determined whether the number of active attackers has decreased. If it has decreased, the defense allocation matrix of the previous time point is adjusted. The corresponding item for the replenished stock attacker is set to 0.

[0114] Based on the above embodiments, if there are still attacking agents that have not been assigned to a defense alliance, such as... Figure 5 As shown, the method also includes:

[0115] Based on a greedy algorithm, each unassigned attacking agent is assigned the nearest defense alliance.

[0116] In this embodiment of the invention, after applying monotonicity constraints, unassigned defenders are assigned to the nearest attacker. This is based on the observation that the closer a defender is to the attacker, the smaller the attacker's safe reach set. The solution of the greedy algorithm is then superimposed onto the assignment matrix. .

[0117] Compared with existing defense allocation methods, this application ensures computational efficiency while effectively guaranteeing continuous improvement and enhancement of defense performance through monotonicity constraints and greedy algorithms.

[0118] Based on the same invention, a defensive allocation device for multi-agent adversarial games is provided. The multi-agent group includes a cluster of defending agents and a cluster of attacking agents. The defending agent cluster consists of several defending agents; the attacking agent cluster consists of several attacking agents. Figure 6 As shown, the device includes:

[0119] Selection unit 601 is used to select several defensive agents from the unassigned defensive agent cluster at the current time to form multiple candidate defense alliances; each candidate defense alliance consists of several defensive agents.

[0120] The search unit 602 is used to search for a target defense alliance among candidate defense alliances based on the constructed zero-sum game objective function, where the attacker aims to maximize the number of attacking agents that invade the target protected area, and the defender aims to minimize the number of attacking agents that successfully invade the target protected area. The target defense alliance and the attacking agents it defends against are used to generate an initial defense allocation strategy. The defense allocation strategy is used to assign a defense alliance to each attacking agent.

[0121] The optimization unit 603 is used to optimize the initial defense allocation strategy generated at the current time using monotonic constraints to obtain the final defense allocation strategy; the monotonic constraint is that the number of attack agents that are expected to be unable to reach the target protection area at the current time increases compared to the previous time.

[0122] Based on the same technical concept, embodiments of the present invention also provide an electronic device, such as... Figure 7 As shown, it includes a processor 701, a communication interface 702, a memory 703, and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 communicate with each other through the communication bus 704.

[0123] Memory 703 is used to store computer programs;

[0124] The processor 701, when executing the program stored in the memory 703, implements the steps of a defense allocation method for multi-agent adversarial games.

[0125] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0126] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0127] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0128] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0129] The computer program product for the defense allocation method for multi-agent adversarial games provided in this embodiment of the invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0130] The defensive allocation device for multi-agent adversarial games provided in this embodiment of the invention can be specific hardware on a device or software or firmware installed on the device. The implementation principle and technical effects of the device provided in this embodiment of the invention are the same as those in the foregoing method embodiments. For the sake of brevity, any parts not mentioned in the device embodiments can be referred to the corresponding content in the foregoing method embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0131] In the embodiments provided by this invention, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0132] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] In addition, the functional units in the embodiments provided by the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0134] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0135] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0136] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. All should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A defense allocation method for multi-agent adversarial games, characterized in that, The multi-agent includes a defense agent cluster and an attack agent cluster, the defense agent cluster is composed of several defense agents; The attack agent cluster is composed of several attack agents; the method includes: Selecting several defense agents in the unassigned defense agent cluster at the current time to form multiple candidate defense alliances; each candidate defense alliance includes several defense agents; the selection process includes: Obtaining the security activity area range of the attack agent cluster and a set of boundary position points of the security activity area range; Selecting the position point closest to the target protection area from the set of boundary position points; Determining the defense agent reaching the closest position point at the same time as the attack agent as the active defense alliance; Selecting a candidate defense alliance in the active defense alliance; Based on the constructed zero-sum game objective function, the attack agent aims to maximize the number of attack agents invading the target protection area, and the defense agent aims to minimize the number of attack agents successfully invading the target protection area, iteratively finding a target defense alliance in the candidate defense alliance, and generating an initial defense allocation strategy based on the target defense alliance and the attack agents it defends; wherein the defense allocation strategy is used to allocate defense alliances for each attack agent; Using a monotonic constraint condition, the initial defense allocation strategy generated at the current time is optimized to obtain the final defense allocation strategy; the monotonic constraint condition is that the number of attack agents expected to be unable to reach the target protection area at the current time increases compared to the last time.

2. The method of claim 1, wherein, Selecting a candidate defense alliance in the active defense alliance includes: In the active defense alliance, selecting an initial active defense alliance, the number of defense agents in the initial active defense alliance is not greater than the dimension of the spatial domain it is in; Selecting the defense alliance with the least number of defenders in the initial active defense alliance and determining it as the candidate alliance.

3. The method of claim 1, wherein, Based on the constructed zero-sum game objective function, the attack agent aims to maximize the number of attack agents invading the target protection area, and the defense agent aims to minimize the number of attack agents successfully invading the target protection area, iteratively finding a target defense alliance in the candidate defense alliance, and generating an initial defense allocation strategy based on the target defense alliance and the attack agents it defends; includes: Iterative optimization based on the constructed zero-sum game objective function; In each iteration process, the attack agent aims to maximize the number of attack agents invading the target protection area, and the defense agent aims to minimize the number of attack agents successfully invading the target protection area, to find a target defense alliance, and generate a defense allocation strategy based on the target defense alliance and the attack agents it defends; Based on the defense allocation matrix obtained by each iteration, a defense allocation matrix is determined; Superimpose the defense allocation matrix obtained by each iteration, and determine the initial defense allocation strategy based on the final superimposed defense allocation matrix.

4. The method of claim 3, wherein, The zero-sum game objective function is: wherein, represents the number of attackers expected to be unable to reach the target protection zone; represents the defense allocation matrix, represents the joint state of all attacking agents and all defending agents; represents the index of the defense coalition; represents the number of defense coalitions; represents the attacking agent index; represents the number of attacking agents; represents the number of defending agents; represents the defending agent index; represents the element of the association matrix.

5. The method of claim 1, wherein, The optimization of the initial defense allocation strategy generated at the current time using a monotonic constraint condition to obtain the final defense allocation strategy includes: predicting, based on an initial defense allocation strategy at the current time, a number of attack agents that cannot reach the target protection region at the current time; determining whether the number of attack agents that cannot reach the target protection region at the current time is greater than the number of attack agents at the previous time; if greater, determining the initial defense allocation strategy at the current time as the final defense allocation strategy at the current time; if less than or equal to, determining the final defense allocation strategy at the previous time as the final defense allocation strategy at the current time.

6. The method of claim 1, wherein, if there are still attack agents that have not been allocated a defense alliance, the method further comprises: allocating, based on a greedy algorithm, each unallocated attack agent to the defense alliance closest to it.

7. A defense allocation apparatus for multi-agent adversarial games, the apparatus comprising: a plurality of agents, each agent having a set of actions; a game state; a set of rules for the game; a set of defense strategies; and a set of defense parameters. The multi-agent includes a defense agent cluster and an attack agent cluster, the defense agent cluster is composed of a plurality of defense agents; The attack agent cluster is composed of a plurality of attack agents; the device comprises: a selection unit configured to select a plurality of defense agents in the unallocated defense agent cluster at the current time to form a plurality of candidate defense alliances; each candidate defense alliance includes a plurality of defense agents; specifically configured to obtain a safe activity region range of the attack agent cluster and a set of boundary position points of the safe activity region range; selecting the position point closest to the target protection region in the set of boundary position points; determining the defense agent that reaches the closest position point at the same time as the attack agent as the proactive defense alliance; selecting a candidate defense alliance in the proactive defense alliance; a finding unit configured to find a target defense alliance in the candidate defense alliance based on a constructed zero-sum game objective function, the attack agent aiming to maximize the number of attack agents that invade the target protection region, and the defense agent aiming to minimize the number of attack agents that successfully invade the target protection region, and generate an initial defense allocation strategy based on the target defense alliance and the attack agents defended thereby; wherein the defense allocation strategy is used to allocate a defense alliance for each attack agent; an optimization unit configured to optimize the initial defense allocation strategy generated at the current time to obtain a final defense allocation strategy using a monotonic constraint condition; the monotonic constraint condition is that the number of attack agents expected to be unable to reach the target protection region at the current time is greater than at the previous time.

8. An electronic device, comprising: comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory is used to store a computer program; the processor is used to execute the program stored in the memory to implement the method steps of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method steps of any one of claims 1-6. The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method steps of any one of claims 1-6.

Images