Non-cooperative drone countermeasures methods and systems based on command and control agents

Abstract

This invention provides a non-cooperative drone countermeasure method and system based on an intelligence agent for command and control, applied in the field of non-cooperative drone countermeasure technology. The method includes: acquiring a counter-drone equipment system; abstracting and encapsulating the counter-drone equipment system based on equipment capabilities and relationships to obtain a temporal network graph; inputting received natural language instructions representing counter-drone intent, the temporal network graph, and a preset adversarial rule base into a preset intelligence agent to obtain a command and control scheme corresponding to the natural language instructions, including: performing task decomposition and task flow orchestration based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain atomic task chains; and performing model matching based on each atomic task in the atomic task chain from a preset professional model library to generate a command and control scheme composed of the atomic task chain and the matched professional model sequence. This invention enables intelligent and automated counter-drone command and control decision-making.

Description

Technical Field

[0001] This invention relates to the field of non-cooperative drone countermeasures technology, and in particular to a non-cooperative drone countermeasures method and system based on a command and control agent. Background Technology

[0002] The widespread proliferation and malicious use of non-cooperative drones (commonly known as "black flight" drones) have seriously threatened the urban environment, critical infrastructure, and social activities.

[0003] The current mainstream command and control model for counter-drone systems is essentially a "human-centric" architecture, with its intelligent construction exhibiting severe fragmentation. Specifically, although some subsystems (such as target recognition and signal analysis) have introduced artificial intelligence (AI) algorithms to improve individual performance, the generation of the entire command and control chain still relies on manual operation. The command center needs to simultaneously monitor multiple unconnected information interfaces, manually completing the fusion of multi-source intelligence, threat level determination, matching of countermeasure rules, and selection and command issuance of interception units.

[0004] This shows that the methods for controlling non-cooperative drones in related technologies suffer from a lack of automation. Summary of the Invention

[0005] This invention provides a method and system for countering non-cooperative drones based on an intelligence agent for command and control, in order to address the shortcomings of existing methods for commanding and controlling non-cooperative drones, which have low levels of automation, and to achieve intelligent and automated anti-drone command and control decision-making.

[0006] This invention provides a non-cooperative drone countermeasure method based on an accusation agent, comprising the following steps: Obtaining an anti-drone equipment system, wherein the anti-drone equipment system includes: drone detection equipment and drone countermeasure equipment; abstracting and encapsulating the anti-drone equipment system based on equipment capabilities and relationships to obtain a temporal network graph; inputting received natural language instructions representing anti-drone countermeasure intentions, the temporal network graph, and a preset adversarial rule base to a preset accusation agent to obtain an accusation scheme corresponding to the natural language instructions output by the accusation agent, wherein the method includes: performing task decomposition and task flow orchestration based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain atomic task chains; performing model matching based on each atomic task in the atomic task chain in a preset professional model library to generate an accusation scheme composed of the atomic task chain and the matched professional model sequence.

[0007] According to the present invention, a non-cooperative drone countermeasure method based on an indictment agent is provided, wherein the anti-drone equipment system is abstracted and encapsulated based on equipment capabilities and equipment relationships to obtain a temporal network graph, including:

[0008] Each equipment node in the anti-drone equipment system is abstracted and encapsulated to form an equipment capability pool. The capability encapsulation information of each equipment node includes: equipment type, capability envelope, concurrent processing capability, and real-time status.

[0009] Based on the real-time connection relationships between various equipment nodes in the equipment capability pool, a time-series complex network is constructed. ,in, This represents the temporal complex network. express The equipment ability pool at all times, express The real-time connection relationships between various equipment nodes in the equipment capability pool at any given moment.

[0010] According to the present invention, a non-cooperative drone countermeasure method based on an indictment agent is provided. The step of decomposing and orchestrating tasks based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain an atomic task chain includes: inputting the natural language instructions, the temporal network graph, and the adversarial rule base into a task parsing and decomposition model based on a large language model to perform constraint-based semantic parsing and task decomposition, obtaining an atomic task set and a constraint set output by the task parsing and decomposition model; inputting the atomic task set and the constraint set into an atomic task flow orchestration model based on a large language model to perform atomic task priority evaluation and task flow orchestration, obtaining an atomic task chain output by the atomic task flow orchestration model.

[0011] According to the present invention, a non-cooperative drone countermeasure method based on an accusation agent includes the following steps: First, performing model matching based on each atomic task in the atomic task chain within a preset professional model library to generate an accusation scheme composed of the atomic task chain and a sequence of matched professional models. This includes: determining the task feature vector of each atomic task in the atomic task chain, wherein the task feature vector includes: task type, solution timeliness requirements, and data size; second, determining the model feature vector of each professional model in the preset professional model library, wherein the model feature vector includes: solvable task types, solution time, solution accuracy, and processable data size; third, determining the matching degree between the task feature vector and the model feature vector based on a weighted similarity function; fourth, performing model adaptation on each atomic task in the atomic task chain based on the matching degree result in the professional model library to obtain a professional model sequence corresponding to the atomic task chain; and fifth, constructing an accusation scheme corresponding to the natural language command based on the atomic task chain and the professional model sequence.

[0012] According to the present invention, a non-cooperative drone countermeasure method based on an accusation agent is provided, wherein the weighted similarity function includes:

[0013]

[0014] in, Indicates the first The atomic task and the first Weighted similarity between professional models Indicates the first Each task feature vector Indicates the first Each model feature vector Represents the weight vector. It represents the Hadamah accumulation. and It is a hyperparameter used to balance the contributions of similarity and difference.

[0015] According to the present invention, a non-cooperative drone countermeasure method based on an intelligence agent for command and control is provided. The method further includes: real-time monitoring of the state changes of each equipment node in the anti-drone equipment system and dynamically updating the time-series network graph; and dynamically optimizing and adjusting the command and control scheme in response to receiving the updated time-series network graph.

[0016] This invention also provides a non-cooperative drone countermeasure system based on an intelligence agent for command and control, comprising the following modules: an acquisition module for acquiring a counter-drone equipment system, wherein the counter-drone equipment system includes: drone detection equipment and drone countermeasure equipment; an encapsulation module for abstracting and encapsulating the counter-drone equipment system based on equipment capabilities and equipment relationships to obtain a temporal network graph; and an intelligence agent module for inputting received natural language instructions representing counter-drone countermeasure intentions, the temporal network graph, and a preset adversarial rule base into a preset intelligence agent to obtain a command and control scheme corresponding to the natural language instructions output by the intelligence agent, wherein the scheme includes: performing task decomposition and task flow orchestration based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain an atomic task chain; and performing model matching based on each atomic task in the atomic task chain in a preset professional model library to generate a command and control scheme composed of the atomic task chain and the matched professional model sequence.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the non-cooperative drone countermeasure method based on the command and control agent as described above.

[0018] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the non-cooperative drone countermeasure method based on a command and control agent as described above.

[0019] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the non-cooperative drone countermeasure method based on a command agent as described above.

[0020] The non-cooperative UAV countermeasure method and system based on command and control agents provided by this invention obtains the anti-UAV equipment system and abstracts and encapsulates it based on equipment capabilities and relationships to obtain a temporal network graph, thereby achieving a unified representation and dynamic relationship modeling of the equipment system. Then, natural language commands, temporal network graphs, and adversarial rule bases are input into the command and control agent. Atomic task chains are obtained through task decomposition and task flow orchestration. Model matching is performed based on atomic tasks in a professional model library to generate a command and control scheme composed of atomic task chains and matched professional model sequences. This significantly improves the intelligence level, ease of operation, and task execution efficiency of anti-UAV command and control decision-making. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the non-cooperative drone countermeasure method based on a command and control agent provided by the present invention.

[0023] Figure 2 This is a schematic diagram of the unified abstract encapsulation of anti-drone equipment capabilities provided by the present invention.

[0024] Figure 3 This is a schematic diagram of the technical route of the command and control intelligent agent provided by the present invention.

[0025] Figure 4 This is a schematic diagram of a module of a non-cooperative drone countermeasure system based on a command and control agent provided by the present invention.

[0026] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0028] The "low, slow, and small" nature of non-cooperative drone targets, with their low detectability, high availability, and low operational threshold, exposes traditional command and control systems based on manual assessment and hierarchical reporting as severely deficient in response delays and decision-making efficiency. Faced with these drone swarms that occur instantaneously, are widely distributed, and may coordinate attacks, building a new generation of counter-drone command and control system capable of real-time, comprehensive perception, intelligent, rapid decision-making, and automated collaborative interception is an urgent and essential requirement.

[0029] The current mainstream command and control model for counter-drone systems suffers from two major flaws: First, decision-making bottlenecks. Human cognitive speed and information processing bandwidth are limited, making them highly susceptible to decision-making delays, fatigue-induced misjudgments, or command conflicts under pressure from complex electromagnetic environments or multiple targets. Second, coordination barriers. The lack of a unified control center for direct and efficient horizontal coordination among detection and strike units means that linkage relies entirely on manual dispatch by commanders, resulting in a sluggish overall system response and difficulty in forming an organic adversarial network. This model of partial "+AI" integration into the command and control system has become a constraint on improving the overall effectiveness of counter-drone systems.

[0030] The necessity of building an AI-native anti-drone command and control agent stems from its fundamental reconstruction of the traditional command and control paradigm. The core positioning of the AI-native anti-drone command and control agent is the autonomous decision-making and collaborative engine of the entire anti-drone system. Its advantages are reflected in three levels: First, in situational awareness, it can achieve deep fusion and real-time calculation of all-source data, automatically constructing a unified and clear situational map, providing a foundation for accurate decision-making. Second, in command and decision-making, its embedded algorithm model can automatically complete threat assessment, resource allocation, and solution generation within milliseconds based on preset adversarial rules and real-time situational awareness, achieving a closed-loop OODA (Observe-Adjust-Decision-Action) cycle of "detect and respond," completely breaking through the speed bottleneck of manual decision-making. Finally, in control and collaboration, the agent, as a unified central hub, can achieve unified management of different equipment and seamless cross-platform and cross-domain collaborative adversarial capabilities, such as automatically guiding radar lock-on, photoelectric tracking and identification, and assigning the optimal jamming station to suppress the drones. Therefore, the AI-native anti-drone command and control agent is an inevitable choice and core pillar for addressing the threats of future intelligent and swarmed drones.

[0031] To address the challenge of countering large-scale non-cooperative drones, this invention provides an AI-native anti-drone command and control intelligent agent framework. This framework features a dedicated large-scale model as its core, various professional AI models as tools, the linkage between the large-scale model and professional models as its hinge, and the unified capability encapsulation and pooling of heterogeneous countermeasure equipment. It enables on-demand registration and access of diverse heterogeneous anti-drone equipment, thorough analysis of highly dynamic confrontation situations, rapid response scheduling of widely distributed anti-drone equipment, and dynamic generation of large-scale command and control links.

[0032] refer to Figure 1 , Figure 1 This is a flowchart illustrating the non-cooperative drone countermeasure method based on a command and control agent provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps.

[0033] Step 101: Obtain the anti-drone equipment system, which includes: drone detection equipment and drone countermeasure equipment.

[0034] In this embodiment of the invention, all available anti-drone equipment needs to be collected and connected to construct an anti-drone equipment system.

[0035] Counter-drone equipment systems are divided into two main categories: drone detection equipment and drone countermeasure equipment. Drone detection equipment includes radar, electro-optical, and radio detection devices, which are responsible for detecting, tracking, and identifying non-cooperative drones; drone countermeasure equipment includes jamming, laser, and net-trapping devices, which are responsible for carrying out soft-kill or hard-kill attacks on targets.

[0036] Step 102: Based on equipment capabilities and equipment relationships, the anti-drone equipment system is abstracted and encapsulated to obtain a time-series network graph.

[0037] In this embodiment of the invention, equipment capability characterization and equipment relationship characterization are performed for each equipment node in the anti-drone equipment system, thereby constructing a time-series network graph.

[0038] A unified abstraction and encapsulation of equipment capabilities is performed for each equipment node in the anti-drone equipment system. This is a standardized description of the inherent attributes of a single equipment node. For any equipment node in the anti-drone equipment system, its equipment capabilities are encapsulated into a structured representation model.

[0039] Based on the encapsulation of equipment capabilities, modeling is performed on the relationships between equipment nodes in the anti-drone equipment system to construct a temporal network graph.

[0040] This invention employs a graph theory-based modeling method to abstract the entire anti-drone equipment system into a dynamic, time-evolving complex network, namely a temporal network graph. In this temporal network graph, each abstractly encapsulated equipment node constitutes a vertex in the network. The relationships between equipment nodes are represented by edges between vertices. Specifically, if two equipment nodes can directly communicate, exchange data, or coordinate in combat at a certain moment, a connecting edge is established between them. This network is temporal, meaning that both the vertex set (the online status of the equipment) and the edge set (the connectivity between equipment) will dynamically update over time, thus realistically reflecting the dynamic situations of equipment entering or leaving the network, malfunctioning, or experiencing link interruptions in the combat environment.

[0041] This involves encapsulating the equipment capabilities of each equipment node and establishing dynamic connections between equipment nodes to abstract and encapsulate the physical anti-drone equipment system into a unified temporal network graph that can be directly processed by a computing system.

[0042] Through the embodiments of the present invention, the temporal network graph not only statically describes the anti-drone equipment in the anti-drone equipment system, but also dynamically represents how the equipment nodes can cooperate, providing an accurate and real-time data foundation for subsequent command and control agents to carry out task planning and resource scheduling.

[0043] Step 103: Input the received natural language instructions, temporal network graph, and preset adversarial rule base used to represent the anti-drone countermeasure intentions into the preset command and control agent to obtain the command and control scheme corresponding to the natural language instructions output by the command and control agent.

[0044] In this embodiment of the invention, the intelligent agent receives natural language instructions (such as "prioritize intercepting the high-speed approaching target on the west side") from the commander (user), and performs automated reasoning and planning by combining the temporal network graph and the pre-set adversarial rule base (including adversarial rules such as "prioritize using laser weapons against large targets").

[0045] Specifically, step 103 includes the following steps.

[0046] Step 1031: Based on natural language instructions, temporal network graphs, and adversarial rule bases, perform task decomposition and task flow orchestration to obtain atomic task chains.

[0047] In this embodiment of the invention, this step is completed collaboratively by a large model within the command agent.

[0048] The Large-Scale Task Parsing and Decomposition Model (LLM-Decomposition) integrates command, graph, and rule base information to analyze the commander's core intent, locate available equipment in the graph, and perform compliance checks. Its output is a set of discretized atomic tasks (such as target identification, threat assessment, and equipment scheduling) and a set of constraints.

[0049] The Large-Scale Atomic Task Flow Orchestration Model (LLM-Planning) prioritizes and links atomic tasks based on the urgency of the adversarial situation and the logical dependencies between tasks, ultimately forming a sequentially executable chain of atomic tasks.

[0050] Step 1032: Perform model matching based on each atomic task in the atomic task chain in the preset professional model library to generate an command and control scheme consisting of the atomic task chain and the matched professional model sequence.

[0051] The command agent will automatically match the optimal solution algorithm for each task in the atomic task chain from a pre-defined professional model library. This library contains various artificial intelligence or operations research models, such as target recognition models, threat assessment models, and resource allocation models. The matching process is implemented through a weighted similarity function, which comprehensively compares the feature vectors of the atomic tasks (such as task type and time requirements) with the characteristic vectors of the algorithm model (professional model) (such as processing speed and accuracy), thereby selecting the most suitable professional model for each atomic task.

[0052] Ultimately, the indictment scheme consists of a pre-ordered chain of atomic tasks and a sequence of specialized models that are matched one by one, forming a complete and automated counter-drone countermeasure process.

[0053] Through the embodiments of this invention, by acquiring the anti-drone equipment system and abstracting and encapsulating it based on equipment capabilities and relationships to obtain a temporal network graph, a unified representation and dynamic relationship modeling of the equipment system is achieved. Furthermore, natural language commands, the temporal network graph, and the adversarial rule base are input into the command and control agent. Atomic task chains are obtained through task decomposition and task flow orchestration. Model matching is performed based on atomic tasks in a professional model library to generate a command and control scheme composed of atomic task chains and matched professional model sequences. This significantly improves the intelligence level, operational convenience, and task execution efficiency of anti-drone command and control decision-making.

[0054] According to the present invention, a non-cooperative drone countermeasure method based on an indictment agent is provided, which abstracts and encapsulates the anti-drone equipment system based on equipment capabilities and equipment relationships to obtain a temporal network graph, including:

[0055] Each equipment node in the anti-drone equipment system has its capabilities abstracted and encapsulated to form an equipment capability pool. The capability encapsulation information of each equipment node includes: equipment type, capability envelope, concurrent processing capability, and real-time status.

[0056] Based on the real-time connection relationships between various equipment nodes in the equipment capability pool, a time-series complex network is constructed. ,in, Representing temporally complex networks, express The equipment ability pool at all times, express The real-time connection relationships between various equipment nodes in the equipment capability pool at any given moment.

[0057] In this invention embodiment, the abstract encapsulation of equipment capabilities is a prerequisite for achieving unified management and control of large-scale heterogeneous countermeasures resources. This invention establishes an equipment representation model based on equipment capabilities and relationships between equipment, and establishes a unified and standardized representation paradigm for detection equipment (radar, photoelectric, electronic reconnaissance, acoustic, etc.) and countermeasure equipment (laser, microwave, jamming, net capture, impact, etc.), forming an equipment capability pool.

[0058] In terms of equipment capability characterization, for any equipment node Its capability encapsulation information can be characterized as ,in Indicates the type of equipment, such as detection equipment, strike equipment, etc.; The capability envelope of the equipment must be specified. For detection equipment, the detection range, identification confidence level, and target accuracy must be clearly defined. For countermeasure equipment, the range envelope, countermeasure style, damage effect, and environmental requirements must be clearly defined. This indicates the equipment's concurrent processing capability. For detection equipment, it is necessary to specify the number of targets that can be detected and tracked simultaneously. For countermeasure equipment, it is necessary to specify whether it is a point-kill or area-kill. It indicates the real-time status of the equipment, such as idle, in a mission, or without mission capability.

[0059] In terms of representing the relationships between equipment, the anti-drone equipment system can be modeled as a temporal complex network based on graph theory methods. ,in Represents the anti-unequipped set (i.e. (Essential equipment pool at all times) Indicates the connection edge between equipment (i.e. (The real-time connection relationships between various equipment nodes in the equipment capability pool at any given moment).

[0060] For example, if equipped and exist If a connection can be maintained at all times, then an edge exists. .

[0061] refer to Figure 2 , Figure 2 This is a schematic diagram of the unified abstract encapsulation of anti-drone equipment capabilities provided by the present invention, which includes multiple targets, multiple detection nodes, command and control nodes, and multiple countermeasure nodes. (These represent different times).

[0062] This invention uses a label graph representation method to visualize and model anti-drone equipment systems (e.g., Figure 2As shown in the figure, all interaction events are added to the constructed static connection edges. In real adversarial scenarios, the link communication time between nodes is very small compared to the node information processing time and can be ignored. Therefore, this invention defines a node interaction event as the subsequent node receiving adversarial information transmitted by the preceding node. The start time of the interaction event is the moment when the preceding node receives the information, and the duration is the time consumed by the preceding node in processing the information.

[0063] In some embodiments, each physical unit with independent function in an anti-drone equipment system, such as a radar, an optoelectronic system, a jamming station, or a command and control unit, is considered an independent equipment node. In implementation, standardized capability description information is defined for each equipment node, thereby encapsulating it into a digital, parameterized entity. The capability encapsulation information mainly includes the following four attributes:

[0064] Equipment Type: This is used to distinguish the fundamental functional category of a node, with a value of "UAV Detection Equipment" or "UAV Countermeasure Equipment". Detection equipment is responsible for target discovery, tracking, and identification, while countermeasure equipment is responsible for implementing soft and hard kill measures against the target.

[0065] Capability envelope: This describes the core performance boundaries of a single piece of equipment using quantifiable parameters. For detection equipment, its capability envelope may include maximum detection range, azimuth and elevation coverage, target identification probability, and data update frequency. For countermeasure equipment, it includes effective range, effective azimuth, countermeasure style (such as suppression jamming and navigation deception), and damage level.

[0066] Concurrent processing capability: This characterizes the upper limit of the number of tasks or targets that a single equipment node can process simultaneously per unit of time. For example, the number of target batches that a phased array radar can track simultaneously, or the number of UAV signal channels that a multi-functional jammer can jam simultaneously.

[0067] Real-time status: Dynamically reflects the availability and working status of equipment nodes at a certain moment. Typical statuses include "ready", "on mission", "faulty", and "offline".

[0068] Through the above encapsulation, all equipment nodes are uniformly abstracted into digital entities with clearly defined capabilities, forming a global, queryable equipment capability pool, which provides a foundation for subsequent capability-based resource matching and mission planning.

[0069] Based on the real-time connection relationships between various equipment nodes in the equipment capability pool, a time-series complex network is constructed. , express The equipment ability pool at any given time, that is, all equipment abilities at any given time. A collection of equipment nodes that are always in an available state, such as "ready" or "on a mission". express The system continuously monitors the real-time connectivity between various equipment nodes in the capability pool. This real-time connectivity is determined by the connectivity of physical communication links, the compatibility of data interface protocols, and the current adversarial grouping rules.

[0070] Through the embodiments of the present invention, by abstracting equipment capabilities and modeling temporal networks, a dynamic and structured representation of the anti-drone equipment system can be achieved, thereby improving the efficiency of collaborative command and control and the real-time response capability against non-cooperative drones.

[0071] According to the present invention, a non-cooperative drone countermeasure method based on an indictment agent is provided, which decomposes and orchestrates tasks based on natural language commands, temporal network graphs, and adversarial rule bases to obtain atomic task chains, including:

[0072] Natural language instructions, temporal network graphs, and adversarial rule bases are input into a task parsing and decomposition model based on a large language model to perform constraint-based semantic parsing and task decomposition, resulting in a set of atomic tasks and a set of constraint conditions output by the task parsing and decomposition model.

[0073] The atomic task set and constraint set are input into the atomic task flow orchestration model based on a large language model to evaluate the atomic task priority and orchestrate the task flow, resulting in the atomic task chain output by the atomic task flow orchestration model.

[0074] In this embodiment of the invention, the purpose of multi-model collaborative reasoning is to transform the commander's qualitative and ambiguous natural language instructions into a set of structured atomic tasks executable by the command and control system. The command and control agent includes: a large-scale task parsing and decomposition model (LLM-Decomposition) and a large-scale atomic task flow orchestration model (LLM-Planning).

[0075] The first step is the context construction of multi-source information fusion, where LLM-Decomposition receives natural language instructions from the commander. The temporal complex network of anti-drone equipment system and the existing adversarial rule base Among them, the adversarial rule base includes adversarial rules such as "large drones should prioritize the use of laser weapons";

[0076] Then, based on constraint-based semantic parsing and task decomposition, the inference process of LLM-Decomposition can be formally expressed as:

[0077] ;

[0078] Among them, the large-scale task parsing and decomposition model needs to parse the commander's natural language instructions. (e.g., "prioritize intercepting targets to the west of the commander") The core intent and focus of action are addressed using an attention mechanism. The lock is applied to the set of equipment used to perform the task, and based on... The compliance and feasibility tests are performed, and the output is a set of discretized atomic tasks. (Such as target intent analysis, threat level assessment, detection equipment scheduling, countermeasure equipment scheduling, time planning, airspace planning, spectrum planning, etc.) and a set of constraints for non-cooperative drone countermeasure operations. (e.g., avoiding civil aviation routes);

[0079] Finally, there is the atomic task priority assessment and task flow orchestration. LLM-Planning analyzes the urgency of the current adversarial situation and combines the input-output chain relationship of atomic tasks to prioritize each atomic task, forming an atomic task chain. It then requests various solution tools in the professional model library to solve various atomic tasks.

[0080] Through the embodiments of the present invention, a large language model is used, combined with natural language instructions, temporal network graphs and adversarial rule bases, to achieve constraint-driven semantic parsing and task decomposition. On this basis, priority evaluation and streaming orchestration of atomic tasks are completed to generate atomic task chains that meet the constraints, thereby improving the intelligent decision-making and task execution capabilities of the command and control system for non-cooperative UAVs.

[0081] According to the present invention, a non-cooperative drone countermeasure method based on an accusation agent is provided, which performs model matching based on each atomic task in an atomic task chain in a preset professional model library to generate an accusation scheme consisting of the atomic task chain and the matched professional model sequence, including:

[0082] Determine the task feature vector for each atomic task in the atomic task chain, where the task feature vector includes: task type, solution timeliness requirements, and data size;

[0083] Determine the model feature vector for each professional model in the preset professional model library. The model feature vector includes: the type of solvable task, the solution time, the solution accuracy, and the scale of data that can be processed.

[0084] Based on the weighted similarity function, the matching degree between the task feature vector and the model feature vector is determined;

[0085] Based on the matching degree results, each atomic task in the atomic task chain is model-adapted in the professional model library to obtain the professional model sequence corresponding to the atomic task chain.

[0086] Based on atomic task chains and professional model sequences, a command and control scheme corresponding to natural language instructions is constructed.

[0087] In this embodiment of the invention, the command and control agent needs to automatically match suitable professional models from a professional model library to solve each atomic task, thereby forming a complete command and control scheme. This invention assumes that various professional models used to solve atomic tasks (such as target intent recognition models, target threat level assessment models, target list ranking models, fire target allocation models, airspace partitioning models, spectrum planning models, etc., which can be constructed using operations research, machine learning, reinforcement learning, etc.) are stored in the professional model library. The core of this invention is to propose an automatic assembly and adaptation technology that automatically pushes suitable models from a pre-set professional model library to each atomic task.

[0088] The first step is to establish a unified task-algorithm feature representation system. For any atomic task... It can be derived from the task feature vector. It means that, among them Indicates the task type of the atomic task. This indicates the timeliness requirements for solving atomic tasks. This indicates the scale of data processed by the atomic task. Each algorithm model (specialized model) in the specialized model library can be represented by a model feature vector. It means that, among them This indicates the types of solvable tasks for the algorithm model (specialized model). This indicates the solution time of the algorithm model. This indicates the solution accuracy of the algorithm model. This indicates the scale of data that the algorithm model can process.

[0089] Next comes the weighted similarity calculation and algorithm matching. Since a single similarity calculation cannot handle the differences in the importance of features across different scenarios, this invention introduces a weighted similarity function. This function combines cosine similarity, which measures directional consistency, and Euclidean distance, which measures absolute distance, to comprehensively evaluate the matching degree between the task and the algorithm.

[0090] Through the embodiments of the present invention, by weighted matching of task features and model features, the atomic tasks and professional models are accurately matched, generating a command and control scheme consisting of an atomic task chain and its corresponding professional model sequence, thereby improving the command and control system's ability to automate and perform high-precision modeling and execution of non-cooperative UAV tasks.

[0091] According to the present invention, a non-cooperative drone countermeasure method based on an accusation agent is provided, wherein the weighted similarity function includes:

[0092]

[0093] in, Indicates the first The atomic task and the first Weighted similarity between professional models Indicates the first Each task feature vector Indicates the first Each model feature vector Represents the weight vector. It represents the Hadamah accumulation. and It is a hyperparameter used to balance the contributions of similarity and difference.

[0094] In an embodiment of the present invention, , and Specific values ​​can be determined through extensive simulation and analysis in a simulation environment. Thus, a complete command and control scheme can be formed by solving each atomic task.

[0095] In practice, the weighted similarity function is used to accurately quantify the matching degree between the atomic task and the professional model, aiming to comprehensively evaluate their similarity and differences. The calculation of this function is based on two core parts: one part measures the cosine similarity in direction between the weighted task feature vector and the model feature vector, reflecting the overall consistency between the two in terms of capability feature dimensions; the other part calculates the weighted distance between the two weighted feature vectors as a penalty term for the degree of difference between them in specific numerical values.

[0096] To balance the contributions of these two components to the final similarity score, an adjustable hyperparameter is introduced. and .in, It is the coefficient of the similarity term. This is the coefficient of the difference penalty term. By adjusting the values ​​of these two hyperparameters, the emphasis of the similarity measure can be controlled; for example, whether more emphasis is placed on the consistency of the overall trend or on the closeness of specific parameters.

[0097] The higher the weighted similarity score, the better the number of similarities. The atomic task and the first The better the matching degree of each specialized model, the better. The accusation agent uses this weighted similarity function to traverse and calculate in the specialized model library for each task in the atomic task chain, selects the specialized model with the highest score, and thus completes accurate model matching, forming the final accusation scheme composed of the atomic task chain and the matched specialized model sequence.

[0098] Through the embodiments of the present invention, the weighted similarity function integrates the cosine similarity and Euclidean distance between the task feature vector and the model feature vector. By balancing the accuracy and difference penalty of the matching with learnable weights and hyperparameters, it achieves refined and adaptive matching between atomic tasks and professional models, thereby improving the accuracy and adaptation efficiency of command and control scheme generation.

[0099] According to the present invention, a non-cooperative drone countermeasure method based on a command and control agent is provided, the method further includes:

[0100] Real-time monitoring of the status changes of each equipment node in the anti-drone equipment system, and dynamic updating of the time-series network graph;

[0101] In response to receiving the updated temporal network graph, the command and control scheme is dynamically optimized and adjusted.

[0102] In this embodiment of the invention, the status changes of each equipment node in the anti-drone equipment system are monitored in real time, and the time-series network graph is dynamically updated.

[0103] Specifically, the system continuously acquires real-time status information reported by each equipment node in the anti-drone equipment system. This information includes, but is not limited to, the equipment's online / offline status, operating mode, remaining resources (such as battery power and ammunition), fault alarms, and the quality of communication links with other nodes. Once a change in the status of any equipment node is detected—for example, a detection node goes offline due to a fault, a countermeasure node completes its current task and transitions to a ready state, or the collaborative link between nodes is interrupted—the temporal network graph update mechanism is triggered. This update directly affects the temporal complex network representing the entire system architecture. The node set is added or removed based on equipment availability, such as removing offline nodes or adding new online nodes; the edge set is reconstructed based on the actual connection relationships between nodes, such as removing interrupted communication links or establishing new logical connections. Through this process, the temporal network graph continuously and synchronously reflects the latest topology and capability distribution of the anti-drone equipment system in combat, providing accurate real-time basis for decision-making.

[0104] When the temporal network graph is updated due to changes in equipment status, this event is transmitted as a trigger signal to the command and control agent. The command and control agent takes the latest temporal network graph, the currently executing or pending atomic task chains, and the pre-set adversarial rule base as input to re-evaluate its decision. The core logic of its dynamic optimization adjustment is to check whether the execution basis of the original command and control scheme has changed due to the graph update (such as the failure of key equipment or the addition of resources).

[0105] For example, if a countermeasure node assigned to perform the jamming task in the original plan suddenly fails, a usable and capable replacement node needs to be reassigned to that atomic task. If a high-performance detection node is added, the task flow may be optimized by reallocating some detection tasks to improve overall efficiency. Optimization and adjustments may involve the partial restructuring of atomic task chains, the rematching of task-resource mapping relationships, or, if necessary, the replanning of some task chains. Ultimately, the command and control agent outputs a new command and control scheme based on the latest system state optimization and issues an execution order, thereby ensuring that the entire counter-drone operation can adapt to dynamically changing adversarial conditions and maintain the effectiveness and robustness of the command and control scheme.

[0106] refer to Figure 3 , Figure 3 This is a schematic diagram of the technical route of the command and control intelligent agent provided by the present invention.

[0107] The commander inputs natural language commands to the command and control agent. The agent first performs multi-model collaborative reasoning task planning, that is, through the LLM-Decomposition module, combined with the adversarial rule base and temporal network graph, performs constraint-based semantic parsing and task decomposition of the commands, outputting a set of atomic tasks and a set of constraints. Then, it is sent to the LLM-Planning module for atomic task priority evaluation and atomic task orchestration, generating an atomic task chain. At the same time, the equipment capability pool abstracts and encapsulates the capabilities of detection and countermeasure equipment and constructs a temporal network graph. Next, in the professional model library, professional models are adapted based on each atomic task in the atomic task chain to generate a command and control scheme consisting of the atomic task chain and the matching professional model sequence. Finally, the commander confirms the command and control scheme and drives the equipment capability pool to execute the corresponding tasks.

[0108] The non-cooperative drone countermeasure system based on the command and control agent provided by the present invention is described below. The non-cooperative drone countermeasure system based on the command and control agent described below can be referred to in correspondence with the non-cooperative drone countermeasure method based on the command and control agent described above.

[0109] refer to Figure 4 , Figure 4 This is a schematic diagram of a module of a non-cooperative drone countermeasure system based on a command and control agent provided by the present invention.

[0110] The acquisition module 401 is used to acquire the anti-drone equipment system, which includes: drone detection equipment and drone countermeasure equipment.

[0111] Encapsulation module 402 is used to abstract and encapsulate the anti-drone equipment system based on equipment capabilities and equipment relationships to obtain a time-series network graph.

[0112] The intelligent agent module 403 is used to input the received natural language instructions representing anti-drone countermeasure intentions, a temporal network graph, and a preset adversarial rule base into a preset command and control intelligent agent to obtain the command and control scheme corresponding to the natural language instructions output by the command and control intelligent agent, which includes:

[0113] Task decomposition and task flow orchestration are performed based on natural language instructions, temporal network graphs, and adversarial rule bases to obtain atomic task chains.

[0114] Based on each atomic task in the atomic task chain, a model matching is performed in the preset professional model library to generate an command and control scheme consisting of the atomic task chain and the matched professional model sequence.

[0115] Specifically, the non-cooperative drone countermeasure system based on the command and control agent provided by the present invention can implement all the method steps implemented in the above-mentioned non-cooperative drone countermeasure method embodiment based on the command and control agent, and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail.

[0116] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communications bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other through the communications bus 540. The processor 510 can call logical instructions in the memory 530 to execute a non-cooperative drone countermeasure method based on an accusation agent. This method includes: acquiring an anti-drone equipment system, which includes drone detection equipment and drone countermeasure equipment; abstracting and encapsulating the anti-drone equipment system based on equipment capabilities and relationships to obtain a temporal network graph; inputting received natural language instructions representing anti-drone countermeasure intent, the temporal network graph, and a preset adversarial rule base to a preset accusation agent to obtain an accusation scheme corresponding to the natural language instructions output by the accusation agent, including: performing task decomposition and task flow orchestration based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain atomic task chains; and performing model matching based on each atomic task in the atomic task chain in a preset professional model library to generate an accusation scheme composed of the atomic task chain and the matched professional model sequence.

[0117] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part 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 of the various embodiments of the present 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.

[0118] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the non-cooperative UAV countermeasure method based on the command and control agent provided by the above methods. The method includes: acquiring an anti-UAV equipment system, wherein the anti-UAV equipment system includes: UAV detection equipment and UAV countermeasure equipment; abstracting and encapsulating the anti-UAV equipment system based on equipment capabilities and equipment relationships to obtain a temporal network graph; inputting the received natural language instructions representing the anti-UAV countermeasure intent, the temporal network graph, and a preset adversarial rule base to a preset command and control agent to obtain a command and control scheme corresponding to the natural language instructions output by the command and control agent, wherein the method includes: performing task decomposition and task flow arrangement based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain an atomic task chain; performing model matching based on each atomic task in the atomic task chain in a preset professional model library to generate a command and control scheme composed of the atomic task chain and the matched professional model sequence.

[0119] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the non-cooperative drone countermeasure method based on the command and control agent provided by the above methods. The method includes: acquiring a counter-drone equipment system, wherein the counter-drone equipment system includes: drone detection equipment and drone countermeasure equipment; abstracting and encapsulating the counter-drone equipment system based on equipment capabilities and equipment relationships to obtain a temporal network graph; inputting the received natural language instructions representing the counter-drone countermeasure intent, the temporal network graph, and a preset adversarial rule base to a preset command and control agent to obtain a command and control scheme corresponding to the natural language instructions output by the command and control agent, wherein the method includes: performing task decomposition and task flow orchestration based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain an atomic task chain; performing model matching based on each atomic task in the atomic task chain in a preset professional model library to generate a command and control scheme composed of the atomic task chain and the matched professional model sequence.

[0120] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0121] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications 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.

Claims

1. A method for countering non-cooperative drones based on a command and control agent, characterized in that, include: Acquire an anti-drone equipment system, wherein the anti-drone equipment system includes: drone detection equipment and drone countermeasure equipment; Based on the equipment capabilities and equipment relationships, the anti-drone equipment system is abstracted and encapsulated to obtain a time-series network graph. The received natural language command representing the anti-drone countermeasure intent, the temporal network graph, and the preset adversarial rule base are input into a preset command agent to obtain the command scheme corresponding to the natural language command output by the command agent, wherein: Based on the natural language instructions, the temporal network graph, and the adversarial rule base, task decomposition and task flow arrangement are performed to obtain atomic task chains; Model matching is performed on each atomic task in the atomic task chain in the preset professional model library to generate an command and control scheme consisting of the atomic task chain and the matched professional model sequence. The step of performing model matching based on each atomic task in the atomic task chain from a preset professional model library to generate an instruction and control scheme consisting of the atomic task chain and the matched professional model sequence includes: Determine the task feature vector for each atomic task in the atomic task chain, wherein the task feature vector includes: task type, solution timeliness requirements, and data size; Determine the model feature vector of each professional model in the preset professional model library, wherein the model feature vector includes: the type of solvable task, the solution time, the solution accuracy, and the scale of data that can be processed; Based on the weighted similarity function, the matching degree between the task feature vector and the model feature vector is determined; Based on the matching degree result, model adaptation is performed on each atomic task in the atomic task chain in the professional model library to obtain the professional model sequence corresponding to the atomic task chain. Based on the atomic task chain and the professional model sequence, construct the command and control scheme corresponding to the natural language instructions; The weighted similarity function includes: ； in, Indicates the first The atomic task and the first Weighted similarity between professional models Indicates the first Each task feature vector Indicates the first Each model feature vector Represents the weight vector. It represents the Hadamah accumulation. and It is a hyperparameter used to balance the contributions of similarity and difference.

2. The non-cooperative UAV countermeasure method based on the accused agent according to claim 1, wherein, The abstract encapsulation of the anti-drone equipment system based on equipment capabilities and equipment relationships yields a time-series network graph, including: Each equipment node in the anti-drone equipment system is abstracted and encapsulated to form an equipment capability pool. The capability encapsulation information of each equipment node includes: equipment type, capability envelope, concurrent processing capability, and real-time status. Based on the real-time connection relationships between various equipment nodes in the equipment capability pool, a time-series complex network is constructed. ,in, This represents the temporal complex network. express The equipment ability pool at all times, express The real-time connection relationships between various equipment nodes in the equipment capability pool at any given moment.

3. The method for countering non-cooperative drones based on a command and control agent according to claim 1, characterized in that, The process of decomposing and orchestrating tasks based on the natural language instructions, the temporal network graph, and the adversarial rule base to obtain atomic task chains includes: The natural language instructions, the temporal network graph, and the adversarial rule base are input into a task parsing and decomposition model based on a large language model to perform constraint-based semantic parsing and task decomposition, thereby obtaining the set of atomic tasks and the set of constraint conditions output by the task parsing and decomposition model. The atomic task set and the constraint set are input into the atomic task flow orchestration model based on a large language model to evaluate the atomic task priority and orchestrate the task flow, thereby obtaining the atomic task chain output by the atomic task flow orchestration model.

4. The non-cooperative UAV countermeasure method based on the accused agent according to claim 1, wherein, The method further includes: The status changes of each equipment node in the anti-drone equipment system are monitored in real time, and the time-series network graph is updated dynamically. In response to receiving the updated temporal network graph, the command and control scheme is dynamically optimized and adjusted.

5. A non-cooperative unmanned aircraft countermeasure system based on a directed agent, characterized in that, include: The acquisition module is used to acquire anti-drone equipment systems, wherein the anti-drone equipment systems include: drone detection equipment and drone countermeasure equipment; An encapsulation module is used to abstract and encapsulate the anti-drone equipment system based on equipment capabilities and equipment relationships to obtain a time-series network graph. An intelligent agent module is used to input received natural language instructions representing anti-drone countermeasure intentions, the temporal network graph, and a preset adversarial rule base into a preset command and control intelligent agent, and obtain a command and control scheme corresponding to the natural language instructions output by the command and control intelligent agent, wherein the scheme includes: Based on the natural language instructions, the temporal network graph, and the adversarial rule base, task decomposition and task flow arrangement are performed to obtain atomic task chains; Model matching is performed on each atomic task in the atomic task chain in the preset professional model library to generate an command and control scheme consisting of the atomic task chain and the matched professional model sequence. The step of performing model matching based on each atomic task in the atomic task chain from a preset professional model library to generate an instruction and control scheme consisting of the atomic task chain and the matched professional model sequence includes: Determine the task feature vector for each atomic task in the atomic task chain, wherein the task feature vector includes: task type, solution timeliness requirements, and data size; Determine the model feature vector of each professional model in the preset professional model library, wherein the model feature vector includes: the type of solvable task, the solution time, the solution accuracy, and the scale of data that can be processed; Based on the weighted similarity function, the matching degree between the task feature vector and the model feature vector is determined; Based on the matching degree result, model adaptation is performed on each atomic task in the atomic task chain in the professional model library to obtain the professional model sequence corresponding to the atomic task chain. Based on the atomic task chain and the professional model sequence, construct the command and control scheme corresponding to the natural language instructions; The weighted similarity function includes: ； in, Indicates the first The atomic task and the first Weighted similarity between professional models Indicates the first Each task feature vector Indicates the first Each model feature vector Represents the weight vector. It represents the Hadamah accumulation. and It is a hyperparameter used to balance the contributions of similarity and difference.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the non-cooperative drone countermeasure method based on a command agent as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the non-cooperative drone countermeasure method based on the command agent as described in any one of claims 1 to 4.

8. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the non-cooperative drone countermeasure method based on the command agent as described in any one of claims 1 to 4.

Images