A cognitive closed-loop cooperative countermeasure system based on digital source jamming and deception

Abstract

The application relates to the technical field of unmanned aerial vehicle countermeasure technology and discloses a cognitive closed-loop cooperative countermeasure system based on digital source interference and deception. The system comprises one master node and at least two slave nodes, and constitutes a closed-loop cooperative architecture of "perception-cognition-decision-execution-evaluation". The master node identifies the intention of an unmanned aerial vehicle through multi-modal fusion, and generates an adaptive interference strategy based on meta-reinforcement learning. A central control unit coordinates the distributed decision of each slave node according to a game model, and forms a cooperative scheme. The slave nodes realize microsecond-level waveform reconfiguration and precise power control through high-precision synchronization and software-defined radio technology, and transmit interference or deceptive signals. The system performs blind evaluation based on channel state information, and dynamically adjusts the strategy. The application realizes the leap from signal recognition to intention understanding, from single-point static interference to multi-node dynamic cooperation, and from open-loop execution to closed-loop evaluation, and significantly improves the intelligence, cooperative efficiency and actual combat robustness of the countermeasure system.

Description

Technical Field

[0001] This invention relates to the field of drone countermeasures technology and discloses a cognitive closed-loop collaborative countermeasure system based on digital source interference and deception. Background Technology

[0002] With the rapid evolution and widespread adoption of drone technology, its applications in both civilian and specialized fields are becoming increasingly extensive, while simultaneously bringing increasingly severe security challenges. In response, drone countermeasures have evolved from traditional single-source jamming to a collaborative system integrating digital source jamming and deception. Existing technologies can achieve frequency-adjustable and power-controllable digital modulation jamming, and can be combined with navigation deception to achieve effects such as deterrence and forced landing.

[0003] However, existing systems still have significant limitations. First, jamming strategies largely rely on pre-set rule bases, lacking real-time awareness of target signal characteristics and adaptive decision-making capabilities, making them inefficient against drones with new or unknown communication systems. Second, the systems mostly employ a single-drone operation mode, limiting their coverage and jamming angles, easily creating blind spots in complex terrain or against highly maneuverable targets, making it difficult to achieve continuous and accurate route guidance. More critically, existing solutions lack real-time, quantitative effectiveness assessment methods after jamming is launched, only able to perform "open-loop" jamming, unable to confirm whether the jamming is effective and to what extent, leading to uncertainty in countermeasures and making optimization difficult.

[0004] Although existing research has addressed these issues by improving hardware RF specifications or enriching the interference waveform library, the fundamental problems stemming from system architecture and decision-making logic—namely, cognitive deficiencies, insufficient coordination, and the lack of an evaluation closed loop—have not yet been systematically resolved. Therefore, there is an urgent need for an innovative countermeasure system design capable of achieving a complete closed loop from target perception, intelligent decision-making, distributed coordination to effect evaluation, thereby fundamentally improving the intelligence level, collaborative efficiency, and overall reliability of countermeasures. Summary of the Invention

[0005] In view of this, this application provides a cognitive closed-loop collaborative countermeasure system based on digital source jamming and deception to achieve intelligent cognition, dynamic coordination and closed-loop evaluation of target UAVs, effectively overcoming the problems of rigid jamming strategies, single-point combat limitations and lack of effect feedback in the prior art, thereby significantly improving the adaptive capability, collaborative efficiency and overall reliability of the countermeasure system.

[0006] The first aspect of this application provides a cognitive closed-loop collaborative countermeasure system based on digital source interference and deception, the system comprising:

[0007] A master node and at least two slave nodes, wherein the master node and each of the slave nodes are interconnected via a wired or wireless network;

[0008] The master node includes:

[0009] The broadband spectrum sensing module is used to perform broadband radio monitoring of the target airspace, capture and digitize the first radio frequency signal of the target UAV in real time, and continuously collect the second radio frequency signal of the target UAV under interference after the countermeasure action is executed.

[0010] The sensor module is used to perform optical monitoring of the target drone and acquire the video stream of the target drone;

[0011] A cognitive processing unit is connected to the broadband spectrum sensing module and the sensor module;

[0012] The cognitive processing unit is configured as follows:

[0013] The first radio frequency signal is subjected to spectrum analysis to extract electromagnetic features, and the video stream is processed to extract visual features, and trajectory features are generated based on the video stream and / or the first radio frequency signal; the electromagnetic features, visual features and trajectory features are correlated and analyzed through a cross-modal fusion network to output the distribution probability of the target UAV intent category;

[0014] Based on the probability distribution of the intent category, an adaptive interference and deception strategy for the target UAV is generated through a meta-reinforcement learning model;

[0015] By comparing and analyzing the first radio frequency signal and the second radio frequency signal, and extracting the statistical feature changes of their channel state information, a quantitative effect assessment of the countermeasures already executed is generated.

[0016] The central control unit is connected to the cognitive processing unit and each of the slave nodes;

[0017] The central control unit is configured to: determine a coordinated jamming and deception scheme in collaboration with each slave node based on the countermeasure mission, the adaptive jamming and deception strategy, and the specific location of the target UAV; and generate and output control commands to each slave node based on the coordinated jamming and deception scheme.

[0018] Receive the quantitative effect evaluation, and dynamically adjust the collaborative interference and deception scheme and corresponding control commands based on the quantitative effect evaluation;

[0019] The slave nodes include:

[0020] A multi-channel digital source module is connected to the central control unit to receive the control commands and generate and output corresponding interference signals or navigation deception signals according to the control commands.

[0021] Optionally, both master and slave nodes include:

[0022] The BeiDou / GPS dual-mode timing module is used to provide the node with a nanosecond-level absolute UTC time reference.

[0023] The IEEE 1588 precision clock protocol processing unit is used to run a precision clock synchronization protocol between the master node and each of the slave nodes, so that the master node and each of the slave nodes keep time synchronized.

[0024] Wherein, the master node is the master clock source of the precision clock synchronization protocol, and the slave node is the slave clock of the precision clock synchronization protocol;

[0025] Based on the time synchronization, the control commands output by the central control unit include the transmission time defined based on absolute UTC time;

[0026] The slave node compensates for its internal processing delay and the signal propagation delay to the master node based on the transmission time, so as to control the multi-channel digital source module to accurately transmit interference signals or navigation decoy signals at the transmission time. The compensation is calculated using the following formula to obtain the start time of the interference signal or navigation decoy signal:

[0027]

[0028] in, Due to internal processing delay, For signal propagation delay, This refers to the launch time.

[0029] Optionally, the cognitive processing unit is configured to generate the distribution probability of the target drone intent category through the following steps:

[0030] The first radio frequency signal is subjected to spectrum analysis to extract electromagnetic features. The target UAV is locked from the video stream using the YOLOX target detection algorithm. The visual features are extracted using a convolutional neural network. The target UAV's flight trajectory is generated by visual tracking using the video stream and / or by TDOA direction finding using the first radio frequency signal. The trajectory features are extracted based on the flight trajectory.

[0031] The electromagnetic features, visual features, and trajectory features are encoded into three independent feature vector sequences, and these three independent feature vector sequences are input into a cross-modal deep learning fusion network for association analysis, outputting the distribution probability of the target UAV's intent category, specifically:

[0032]

[0033] in, , , These are the query vector, key vector, and value vector, respectively, obtained by linearly transforming the feature vector sequences corresponding to the electromagnetic features, visual features, and trajectory features. is the dimension of the key vector.

[0034] Optionally, the construction and use of meta-reinforcement learning models include the following steps:

[0035] In the pre-training phase, a training task distribution containing multiple UAV communication models is constructed in a simulation environment. Meta-training is performed on all tasks based on a meta-learning algorithm to obtain initial policy model parameters with cross-task transferability. Specifically:

[0036]

[0037] in, These are the initial policy model parameters. Represents the first in the training task distribution One task, The learning rate within the task. Representative to Mathematical operators for finding gradients For the task loss function For parameters gradient, In the mission The adaptation strategy parameters obtained after fine-tuning;

[0038] During the online adaptation phase, when encountering an unknown target drone, the cognitive processing unit defines the current scenario as a new meta-task, loads the initial strategy model parameters θ, and performs rapid gradient updates on the parameters θ through exploratory interactions with the target drone and based on the instantaneous reward signal obtained from the quantitative effect evaluation, to obtain optimized strategy parameters for the current target drone. Based on the optimized strategy parameters, an adaptive interference and deception strategy adapted to the current target drone is generated.

[0039] Optionally, the central control unit is configured to generate and output the control commands through the following steps:

[0040] Based on the countermeasures mission, the adaptive jamming and deception strategy, and the specific location of the target UAV, the central control unit defines the strategy space and utility function for each slave node.

[0041] The utility function for:

[0042]

[0043] in, To determine the probability of a successful counterattack. The energy consumption of the slave node itself. The probability of being detected by the enemy. , , These are the corresponding weighting coefficients;

[0044] Each slave node autonomously iteratively updates its policy based on the utility function using a distributed decision algorithm, which is a log-based linear learning algorithm. At any moment Choose the strategy with the following probabilities :

[0045]

[0046] in, This indicates that during the summation process, the nodes are traversed. strategy space Every possible strategy taken at that time Indicates the strategy of other slave nodes. Parameters for controlling the degree of exploration, For the node The strategy space;

[0047] Through the iteration of the distributed decision-making algorithm, the strategies of each slave node converge to the Nash equilibrium point. The central control unit combines the strategies adopted by each slave node at the equilibrium point and integrates them into the cooperative interference and deception scheme. Based on the cooperative interference and deception scheme, it generates and outputs control commands to each slave node.

[0048] Optionally, the cognitive processing unit is configured to generate the quantitative performance evaluation through the following steps:

[0049] Extract the channel state information of the first radio frequency signal and the second radio frequency signal;

[0050] Calculate the statistical feature change of the channel state information of the second radio frequency signal relative to the channel state information of the first radio frequency signal to construct an interference fingerprint feature vector. The statistical features include at least amplitude perturbation entropy and phase noise power.

[0051] The interference fingerprint feature vector is input into a pre-trained interference effect classifier; the output is a quantitative evaluation result of the countermeasure effect, which is used to characterize the effectiveness level of interference or deception.

[0052] Optionally, the multi-channel digital source module includes a direct digital synthesis unit, a field-programmable gate array, a high-speed digital-to-analog converter, a control unit, and a high-linearity power amplifier link;

[0053] The multi-channel digital source module is configured as follows:

[0054] According to the control command, the corresponding baseband I / Q waveform signal is dynamically generated in the field programmable gate array;

[0055] The baseband I / Q waveform signal is converted into an analog radio frequency signal using the direct digital synthesis unit and high-speed digital-to-analog converter.

[0056] The analog radio frequency signal is amplified by the high linearity power amplification link and output as the interference signal or navigation decoy signal;

[0057] The multi-channel digital source module supports generating radio frequency signals with a frequency range of 100MHz to 6GHz, supports multiple modulation methods including white noise, QPSK, BPSK, and OFDM, and has at least 4 independent signal output channels.

[0058] Optionally, the field-programmable gate array is configured such that its internal signal generation pipeline parameters, including carrier frequency, modulation type, symbol rate, and filter coefficients, can be dynamically reconfigured according to the control commands within microseconds.

[0059] Furthermore, the power control of the multi-channel digital source module adopts a combination of digital predistortion and closed-loop power detection, specifically:

[0060] When generating the baseband I / Q waveform signal, the field-programmable gate array performs digital amplitude predistortion scaling according to the target power value in the control command;

[0061] The output of the high linearity power amplifier link is equipped with a power detection circuit to detect the actual output power and feed it back to the control unit;

[0062] The control unit is configured to perform closed-loop adjustment of the digital amplitude predistortion scaling based on the target power value and the actual output power feedback, so that the transmission power accuracy of the interference signal or navigation decoy signal is better than ±1dB.

[0063] Optionally, when at least two slave nodes need to perform focused jamming on the same distant target UAV,

[0064] The central control unit is also configured to: calculate the signal propagation path difference from each node to the target UAV, and convert the propagation path difference into an accurate transmission time difference or carrier initial phase difference;

[0065] Generate differentiated control commands for each slave node, including the transmission time difference or the initial phase difference of the carrier.

[0066] Each slave node controls the multi-channel digital source module to generate and transmit interference signals or navigation decoy signals according to the differentiated control instructions, so that the interference signals or navigation decoy signals transmitted by multiple slave nodes are spatially superimposed at the location of the target UAV.

[0067] Optionally, the system also includes state synchronization and dynamic fault tolerance mechanisms, specifically:

[0068] Each of the slave nodes is configured to periodically output its own status information data to the central control unit of the master node. The status information data includes at least health status, remaining power capacity, and synchronization lock status.

[0069] The central control unit is configured to monitor the status information data of each slave node in real time. When it is determined that any slave node has failed or has insufficient performance, based on the quantitative effect evaluation and the status of the remaining slave nodes, the determination process of the cooperative interference and deception scheme is re-executed, and the updated control commands are dynamically distributed to other slave nodes.

[0070] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present invention include at least the following:

[0071] By constructing a cognitive closed-loop collaborative architecture of "perception-cognition-decision-execution-evaluation," the system can leap from traditional signal recognition to a deep understanding of the UAV's mission intent. Based on meta-reinforcement learning, it achieves rapid online adaptation of jamming strategies, fundamentally overcoming the limitations of existing technologies that rely on preset rules and cannot cope with unknown targets. This significantly improves the intelligence level of countermeasures and the ability to respond to new threats. Simultaneously, by modeling the collaborative countermeasure process as a dynamic non-cooperative game and employing distributed decision-making algorithms, multiple nodes can autonomously collaborate based on local information to form optimal resource allocation and action plans. Combined with a high-precision hierarchical synchronization network and waveform microsecond-level reconfiguration capabilities, it achieves precise spatiotemporal coordination and energy focusing of multi-node signals at the target, thus completely breaking through the coverage blind spots and efficiency bottlenecks of single-point operations. Furthermore, by innovatively achieving blind evaluation of jamming effects based on changes in channel state information, the system can complete closed-loop feedback and dynamic strategy optimization without demodulating the target signal. It also possesses fault-tolerant capabilities for synchronous monitoring of node states and dynamic mission migration, thereby ensuring the overall high reliability, robustness, and continuous combat effectiveness of the system in complex adversarial environments. Attached Figure Description

[0072] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0073] Figure 1 This is a system configuration diagram of the present invention;

[0074] Figure 2 This is a flowchart of the master-slave node operation of the present invention. Detailed Implementation

[0075] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0076] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0077] Example 1

[0078] As can be seen from the above background technology, in the face of increasingly complex and intelligent drone threats, how to overcome the fundamental defects of existing countermeasure systems, such as static and rigid interference strategies, limited single-point combat capabilities, and lack of real-time effect feedback, in order to achieve efficient, reliable, and intelligent regional air defense, is one of the key technical challenges currently facing the security and countermeasure field.

[0079] In existing technologies, typical integrated digital source jamming and decoy devices can achieve wide-band, multi-mode signal transmission, but their jamming modes (such as forced landing and drive-away) and parameters (frequency, power) largely depend on preset rules or manual selection. When facing drones using new communication protocols, frequency hopping patterns, or unknown modulation methods, such systems cannot recognize target characteristics in real time and dynamically adjust strategies, often leading to jamming failure or energy waste. Furthermore, these devices are mostly independent operational units with limited coverage and beam directionality, making it difficult to form continuous, seamless jamming suppression zones or precise flight path guidance corridors in complex urban environments or when targeting high-speed maneuvering targets. More significantly, after transmitting jamming signals, operators typically cannot obtain quantitative, real-time feedback on whether the jamming is effective and to what extent; the entire countermeasures process is in an "open-loop" state, severely impacting the scientific nature of decision-making and the effectiveness of actions.

[0080] Therefore, this application provides a cognitive closed-loop collaborative countermeasure system based on digital source interference and deception, such as... Figure 1 and Figure 2 As shown, the system includes: a master node and at least two slave nodes, wherein the master node and each of the slave nodes are interconnected via a wired or wireless network and work together.

[0081] The master node is the core command and processing center of the system, and specifically includes the following modules:

[0082] Wideband Spectrum Sensing Module: This module is used for wideband (e.g., 300MHz to 6GHz) radio monitoring of the target airspace. Its core function is to scan, capture, and digitize the first radio frequency signal emitted by the target UAV (i.e., the original signal before interference) in real time, and continuously collect the second radio frequency signal of the target UAV under interference after the system executes countermeasures.

[0083] Sensor module: This module is used for optical monitoring of the target UAV, and is usually an integrated high-definition camera or photoelectric turntable, responsible for acquiring video streams containing the target UAV.

[0084] Cognitive Processing Unit: This unit is hardware-connected to the broadband spectrum sensing module and the sensor module to receive radio frequency signals and video stream data. It is configured to perform three core functions:

[0085] Target signal cognitive analysis (target intent recognition): Spectral analysis is performed on the first radio frequency signal to extract electromagnetic features, while the video stream is processed to extract visual features, and visual tracking is performed based on the video stream to generate trajectory features. Subsequently, a cross-modal fusion network is used to perform correlation analysis on the above three types of features, and finally, the distribution probability of the target UAV intent category is output (e.g., {reconnaissance: 0.8, penetration: 0.15, delivery: 0.05}).

[0086] Adaptive jamming strategy generation: Based on the probability distribution of the intent category, an adaptive jamming deception strategy for the target UAV is generated by calculating through a meta-reinforcement learning model. This strategy includes the suggested jamming pattern, operating frequency band, power level, etc.

[0087] Interference effect closed-loop evaluation: By comparing and analyzing the first and second radio frequency signals before and after interference, and extracting the statistical feature changes of their channel state information, a quantitative evaluation of the effect of the countermeasures already implemented is generated.

[0088] Central Control Unit: This unit is connected to the cognitive processing unit and each of the slave nodes. It is configured for two core functions:

[0089] Collaborative task planning and allocation: Based on the countermeasures task (such as driving away) issued by the operator, the adaptive interference and deception strategy, and the specific location of the target UAV, the collaborative interference and deception scheme is determined by each slave node, and specific control commands are generated based on this scheme and distributed to each slave node through the network.

[0090] Closed-loop control: Receives quantitative effect evaluation from the cognitive processing unit and dynamically adjusts subsequent coordinated interference and deception schemes and corresponding control commands based on the evaluation results to optimize subsequent countermeasures.

[0091] The slave node is the system's distributed signal execution unit, specifically including:

[0092] Multi-channel digital source module: This module is connected to the central control unit of the master node for communication, and is used to receive control commands issued by it. The main core function of this module is:

[0093] High-precision waveform synthesis: Accurately generates and outputs corresponding interference signals or navigation decoy signals (waveform signals) according to command requirements. Its hardware typically includes a field-programmable gate array (FPGA), a high-speed digital-to-analog converter (DAC), and a high-linearity power amplifier link, supporting the generation of signals with various modulation styles within the 100MHz to 6GHz frequency band.

[0094] Cooperative control: Each slave node compensates for the time delay of the transmission time specified in the command based on the unified high-precision time reference of the system, and transmits the signal at a precise time of strict synchronization; when multiple slave nodes act on the same target, the system controls the transmission time or phase to make the signals spatially in phase superimposed at the target position to enhance the interference effect.

[0095] Ultimately, each slave node works together to transmit interference or decoy signals.

[0096] System Workflow

[0097] The system's workflow within a complete countermeasure cycle is as follows:

[0098] Sensing and Signal Acquisition: After system deployment, the broadband spectrum sensing module of the master node continuously scans the airspace. When a suspicious UAV target is detected, it immediately locks onto and digitizes its first radio frequency signal. At the same time, the sensor module adjusts its pointing to capture the target's video stream.

[0099] Target signal cognitive analysis: The cognitive processing unit simultaneously processes radio frequency and video data. It analyzes signal characteristics to identify target models; analyzes flight trajectories to determine their movement patterns; and finally, through a cross-modal fusion network, it comprehensively determines the target's intent (e.g., identifying it as a drone conducting area reconnaissance).

[0100] Adaptive jamming strategy generation: Based on the "reconnaissance" intent, the cognitive processing unit quickly generates an adaptive jamming and deception strategy for the purpose of "driving away" the target through its meta-reinforcement learning model, suggesting prioritizing jamming of the target's navigation link and implementing heading deception. Upon receiving the strategy, the central control unit coordinates the two slave nodes based on the target's real-time position. For example, it decides that slave node A is responsible for transmitting suppressive jamming signals, while slave node B is responsible for transmitting navigation deception signals, and calculates the temporal and spatial relationship of their coordinated transmissions to form a coordinated jamming and deception scheme.

[0101] Collaborative task planning and allocation: The central control unit converts the above scheme into control commands containing specific frequency, power, modulation method and absolute transmission time, and sends them to slave nodes A and B respectively.

[0102] High-precision waveform synthesis and collaborative execution:

[0103] Upon receiving control commands from nodes A and B, their internal multi-channel digital source modules immediately activate. The field-programmable gate array (FPGA) dynamically generates corresponding baseband I / Q waveforms based on the commands, which are then synthesized into the specified interference or navigation decoy RF signals via high-speed digital-to-analog conversion and a high-linearity power amplification link. During this process, each node compensates for the transmission time specified in the command based on the system's unified high-precision time reference, achieving microsecond-level synchronization. When multiple nodes coordinate their actions against the same target, the system precisely controls the transmission time or initial carrier phase of each node, enabling the signals to be spatially in-phase superimposed at the target UAV, thus achieving focused interference or coordinated decoy effects.

[0104] Interference effect closed-loop evaluation: Simultaneously with signal transmission, the master node's broadband spectrum sensing module continuously acquires the second radio frequency signal after the target has been interfered with. The cognitive processing unit immediately initiates the evaluation process, extracts and compares the channel state information of the first and second radio frequency signals, and generates a quantitative effect evaluation (e.g., "navigation signal quality decreased by 90%)" by calculating the change in characteristics. This evaluation is fed back to the central control unit in real time.

[0105] Closed-loop control: If the evaluation results show that the interference effect is unsatisfactory, the central control unit will immediately trigger an adjustment mechanism. For example, it will instruct the cognitive processing unit to update the interference strategy, or re-coordinate the changes in interference parameters from slave nodes, and start a new round of "command-execution-evaluation" cycle until the drone is successfully driven away. After the mission is completed, the system resets and waits for the next threat.

[0106] Example 2

[0107] This embodiment aims to further describe a high-level, preferred implementation of the system of the present invention, and illustrates the system scenario and initialization:

[0108] Suppose a high-end industrial drone, employing an unknown proprietary frequency-hopping protocol and encrypted image transmission, appears in an urban area with complex electromagnetic background noise. Its flight trajectory exhibits evasive maneuvers. The system objective is to accurately locate and "deceive" the drone for capture. The system deployment is as follows: Figure 1 As shown, it includes one master node and three slave nodes (numbered N1, N2, N3), which are interconnected through a low-latency wireless mesh network.

[0109] After all nodes are powered on, the built-in BeiDou / GPS dual-mode timing module provides each node with an absolute UTC time reference with nanosecond-level accuracy. Subsequently, the IEEE 1588 precision clock protocol processing units of the master node and all slave nodes are activated. The master node acts as the master clock source, and the slave nodes act as slave clocks. By exchanging synchronization messages with hardware timestamps, the entire network achieves sub-microsecond-level time synchronization. This step provides a unified, high-precision time reference for all subsequent operations requiring strict timing coordination, and is the fundamental prerequisite for achieving precise synchronization and even in-phase superposition of multi-node signals at spatial target points. This demonstrates the system's complete workflow and superior performance in complex adversarial scenarios.

[0110] Refined Feature Extraction (Target Signal Cognitive Analysis)

[0111] Electromagnetic characteristics Extraction: The cognitive processing unit performs real-time spectrum analysis, demodulation identification, and blind signal analysis on the first radio frequency signal. Within seconds, it identifies that the signal uses a non-standard frequency hopping pattern (e.g., hopping rate of 100 hops / s, frequency set unknown) and extracts electromagnetic characteristics such as instantaneous bandwidth and symbol rate. Beneficial effects: Beyond simple signal detection, it enables in-depth analysis of complex and unknown communication systems.

[0112] Visual features Extraction: The cognitive processing unit invokes the deployed YOLOX target detection algorithm to quickly and accurately frame the drone target in each frame of the video stream. Subsequently, a lightweight convolutional neural network is used to extract deep features from the target area image, outputting a visual feature vector containing the drone model, size, and key morphological features (such as whether it is equipped with a gimbal or special payload). Beneficial effects: Upgrading optical observation from "seeing the target" to "identifying the target model and mission configuration" provides high-level semantic information.

[0113] Trajectory features Extraction: The system integrates two methods to generate accurate flight paths: First, based on continuous frames of the video stream, a visual tracking algorithm (such as KCF) is used to generate the target's planar trajectory, and the 3D position is calculated by combining the pod attitude data. Second, using the master node or its auxiliary distributed radio frequency sensing array, TDOA direction finding is performed on the first radio frequency signal to obtain another set of 3D position estimates. These two sets of data are then fused using a Kalman filter to generate a smooth and accurate sequence of UAV 3D flight path points. This allows for the calculation of trajectory characteristics such as instantaneous velocity, acceleration, and rate of change of heading angle. Beneficial effects: It provides high-precision target motion state information, serving as a crucial input for determining tactical intentions.

[0114] Cross-modal intent recognition (target signal cognitive analysis)

[0115] The cognitive processing unit inputs the three sets of features into the encoder and converts them into three feature vector sequences of the same dimension.

[0116] These sequences are fed into a cross-modal attention fusion network based on the Transformer architecture. This network employs a multi-head self-attention mechanism. Calculations are performed, in which, , , These are the query vector, key vector, and value vector, respectively, obtained by linearly transforming the feature vector sequences corresponding to the electromagnetic features, visual features, and trajectory features. is the dimension of the key vector.

[0117] The network automatically learns the association weights between modalities, for example, when Identified "equipped with a high-resolution camera" and When displaying "low-speed gridded flight", the network assigns higher attention weights to these feature combinations.

[0118] The fused features are fed into a fully connected classifier, which ultimately outputs the probability distribution of the target UAV's intent category, for example {fine mapping: 0.75, surveillance and reconnaissance: 0.20, others: 0.05}. This achieves a mapping from low-level signal features to high-level mission intent, enabling the system to "understand" what the target is doing, thus providing unprecedented decision-making basis for implementing targeted and differentiated countermeasure strategies.

[0119] Meta-reinforcement learning-driven adaptive policy generation (adaptive perturbation policy generation)

[0120] Policy Model Initialization: The cognitive processing unit loads a meta-trained meta-reinforcement learning model. The initial policy model parameters θ are obtained through meta-learning (such as the MAML algorithm) on a "task distribution" consisting of a large number of known and unknown UAV communication models in a simulation environment. The meta-training process follows the formula:

[0121]

[0122] in, These are the initial policy model parameters. Represents the first in the training task distribution One task, The learning rate within the task. Representative to Mathematical operators for finding gradients For the task loss function For parameters gradient, In the mission The adaptation strategy parameters obtained after fine-tuning;

[0123] Its goal is to minimize the sum of the losses of the adapted policies on all tasks, so that the model has the ability to "learn to learn", that is, to adapt quickly when faced with new tasks.

[0124] Online rapid adaptation and strategy generation: The drone currently being encountered, which uses an unknown frequency hopping protocol, is defined by the system as a new meta-task. The system does not rely on any preset rule base, but instead initiates an online adaptation process:

[0125] Control a slave node (e.g., N1) to sequentially transmit three low-power, different types of probing interference (e.g., short-time broadband noise, fixed-frequency QPSK interference, and frequency sweeping interference).

[0126] After each trial, the master node immediately generates an instant reward signal r through the subsequent evaluation module (for example, r is positive if the jitter of the target signal increases after interference, and negative if there is no change).

[0127] Using this limited amount of (state, action, reward) interaction data, the meta-model parameters θ are updated 1-2 times using gradients to obtain the optimized policy parameters for the current objective. .

[0128] based on The cognitive processing unit generates specific adaptive jamming and decoy strategies, such as: "implementing predictive guided frequency targeting jamming on the frequency hopping control link, using noise-like QPSK modulation of the jamming waveform to maximize the symbol error rate; implementing partial frequency band blocking jamming on the image transmission link to induce it to switch to an easily jammed frequency band; and simultaneously generating a virtual GPS navigation decoy route to a preset safe recovery point." This endows the system with the ability to "learn and fight immediately" in response to unknown threats, completely eliminating reliance on prior knowledge bases, achieving dynamic, autonomous, and rapid optimization of countermeasure strategies, and maintaining the long-term effectiveness of the technology.

[0129] Game theory-driven distributed collaborative decision-making (collaborative task planning and allocation)

[0130] Constructing the game model: Based on the "deception and capture" task, the aforementioned strategies, and the target's real-time location, the central control unit models the three slave nodes as three agents (players) in a dynamic non-cooperative game. A policy space S_i is defined for each node i, for example, S_i = {interference control link, interference image transmission link, transmit deception signal, coordinated interference, standby}.

[0131] Defining Utility and Distributed Solution: Designing a utility function that conforms to the properties of potential games. :

[0132]

[0133] in, To determine the probability of a successful counterattack. The energy consumption of the slave node itself. The probability of being detected by the enemy. , , These are the corresponding weighting coefficients;

[0134] in, The success probability depends on the joint strategy of all nodes, reflecting a synergistic effect. Each node runs a log linear learning algorithm in parallel based on local information (its own position, battery level, and relative target location). At time t, each node i has a probability... Choose strategy s, where This indicates that during the summation process, the nodes are traversed. strategy space Every possible strategy taken at that time Indicates the strategy of other slave nodes. Parameters for controlling the degree of exploration, For the node The strategy space;

[0135] Through repeated iterations and the exchange of limited neighboring node information, the entire system spontaneously converges to a pure policy Nash equilibrium.

[0136] Dynamic Alliance Formation and Scheme Integration: In this scenario, after game convergence, the system automatically forms two "dynamic alliances": nodes N1 and N2 form an alliance, jointly responsible for coordinated tracking interference of the frequency-hopping link (to improve reliability and form interference sectors); node N3 focuses on generating and transmitting highly realistic navigation decoy signals. The central control unit combines and integrates this equilibrium strategy into a coordinated interference and decoy scheme. Beneficial Effects: Decentralized and highly robust resource allocation is achieved. The addition, removal, or failure of a single node does not cause global planning to collapse; the system can automatically re-achieve optimal coordination through distributed computing, greatly improving the system's survivability and adaptability in complex adversarial environments. Game Theory-Driven Distributed Cooperative Decision-Making

[0137] High-precision synchronization and agile waveform execution (high-precision waveform synthesis and collaborative execution)

[0138] Command generation and distribution: The central control unit translates the above scheme into specific control commands. These commands include not only waveform parameters and power, but most importantly, the transmission time based on absolute UTC time. .

[0139] Agile waveform synthesis: Instructions are received from the multi-channel digital source module of the node. Its core field-programmable gate array (FPGA) dynamically reconfigures parameters such as carrier frequency, modulation type (e.g., switching from QPSK to OFDM), symbol rate, and filter coefficients of the signal generation pipeline within microseconds, based on the instructions. For example, to track frequency hopping, the FPGA can switch the frequency from f1 to f2 in the next clock cycle based on the predicted frequency hopping pattern. Beneficial effects: It enables software definition and millisecond-level agile transformation of interference waveforms, allowing interference signals to precisely match and counteract the communication characteristics of the target, much like a "surgical scalpel."

[0140] Precision Power Control: The power control loop operates synchronously. The FPGA performs digital amplitude pre-distortion scaling in the digital domain based on the target power value; the amplified RF signal is sampled by the power detection circuit at the power amplifier output, and the actual power value is fed back to the control unit; the control unit compares the target value with the actual value, dynamically adjusts the pre-distortion parameters, forming a closed loop to ensure the stability and accuracy of the output power at any frequency (better than ±1dB). Beneficial Effects: It achieves refined control of interference power, ensuring effective interference while avoiding excessive emission that exposes itself or causes unnecessary electromagnetic pollution.

[0141] Synchronous launch and space focusing: Each slave node launches according to the absolute launch time in the command. And according to the formula:

[0142]

[0143] Calculate the local startup time. Among them, It is a known fixed processing delay within the node. This involves calculating the signal propagation delay based on the theoretical positions of the nodes and the target (with pre-compensation). For nodes N1 and N2, which require spatial focusing, the central control unit pre-calculates the path difference between them and the target point. And convert it into a precise launch time difference. (where c is the speed of light), and this is reflected in their respective control commands. Ultimately, N1 and N2 are in phase difference. The system transmits coherent jamming signals at precise moments, causing them to be spatially superimposed at the target UAV receiver. Beneficial effects: It improves time synchronization accuracy from "simultaneous transmission" to "synchronous arrival at the target point," achieving spatial energy convergence. Without increasing the power of individual nodes, it significantly increases the effective jamming power at the target, while reducing the risk of individual detection and positioning of each node's transmitted signals.

[0144] Blind evaluation based on channel state information (CSI) (closed-loop evaluation of interference effects)

[0145] Data Acquisition and Feature Extraction: Simultaneously with the jamming transmission, the master node's sensing array continuously receives the target signal (second radio frequency signal). The cognitive processing unit utilizes its multi-antenna capability to accurately estimate the channel frequency response of the target signal before and after jamming, i.e., channel state information (CSI), denoted as... and .

[0146] Constructing an interference fingerprint: Calculate the statistical characteristic change of the CSI of the second radio frequency signal relative to the CSI of the first radio frequency signal. For example:

[0147] Calculate the amplitude perturbation entropy: Interference causes the amplitude distribution of the multipath channel to become more random, and the entropy value increases.

[0148] Calculate phase noise power: Effective interference will introduce significant phase jitter.

[0149] These features are used to construct the perturbation fingerprint feature vector. .

[0150] Amplitude perturbation entropy :

[0151] Physical meaning: It measures the degree of disorder or randomness in the amplitude values ​​of the channel frequency response (CSI). In wireless communication, the distribution of the received amplitude of a normal signal under multipath propagation usually exhibits certain statistical regularities. When an effective interference signal is injected, it interacts in complex ways with the original signal, disrupting the original channel characteristics and causing the fluctuations in the received signal amplitude to become more disordered and unpredictable.

[0152] Mathematical processing: Calculate the information entropy of the probability distribution of amplitude values. The higher the entropy value, the more random and chaotic the amplitude distribution.

[0153] System objective: A significant increase in amplitude is a key indicator that interference energy is disrupting signal amplitude stability. This typically means that the target receiver is struggling to accurately estimate and compensate for channel variations, leading to a decline in demodulation performance.

[0154] Phase noise power :

[0155] Physical meaning: It measures the phase jitter or noise energy in the channel frequency response. Phase information is crucial for demodulation in many modern communication modulation methods (such as QPSK, QAM), requiring a high degree of consistency and stability. Interference introduces additional phase perturbations, disrupting this consistency.

[0156] Mathematical processing: This is typically obtained by analyzing the fluctuations in the phase sequence (e.g., calculating the noise power in its spectrum at the offset carrier frequency).

[0157] System objective: An increase in phase noise directly reflects the disruption of signal phase integrity caused by interference. Strong phase noise can lead to inter-symbol interference and a spike in the bit error rate in the receiver, which is an important indicator of communication link interruption or severe degradation.

[0158] Categorized assessment: Input an SVM classifier pre-trained on a large amount of experimental data. The classifier outputs a quantized performance evaluation result, for example: {Complete control link blockage: 0.88, Severe image transmission link degradation: 0.75, Decoy signal not received: 0.10}. This innovatively achieves "knowing the effect without decoding." Even if the target signal uses encryption or a proprietary protocol, the system can passively and quantitatively evaluate the interference effect through changes in physical layer channel characteristics, providing a reliable and universal feedback signal for closed-loop control.

[0159] Closed-loop dynamic adjustment and system fault tolerance (closed-loop control)

[0160] Policy closed-loop optimization: The above-mentioned quantitative effect evaluation is fed back to the central control unit and cognitive processing unit in real time. If the evaluation shows that "the decoy signal was not received", the central control unit may instruct the adjustment of the power of the decoy signal or the encoding strategy; at the same time, the evaluation is fed back as a "reward" signal to the meta-reinforcement learning model for further fine-tuning of the online policy parameters.

[0161] State synchronization and dynamic fault tolerance: All slave nodes periodically report their status information data to the master node, including health status, remaining power capacity, synchronization lock status, etc. The central control unit provides global monitoring. In this embodiment, it is assumed that node N2 suddenly fails due to a hardware fault. The central control unit immediately detects this fault through a heartbeat loss mechanism.

[0162] Dynamic Mission Migration: Once N2 is determined to have failed, the central control unit, based on the latest quantified effect assessment (which may show a decrease in interference effectiveness) and the status of the remaining nodes N1 and N3, immediately re-executes the process of determining the coordinated interference and decoy scheme. In the new game calculation, the system may decide that N1 will independently undertake the interference to the control link (or adjust the interference pattern to compensate for power loss), and generate updated control commands to dynamically allocate to N1 and N3. The entire adjustment process can be completed within seconds. Beneficial Effects: It endows the system with strong resilience and adaptability. The failure of a single node will not lead to mission failure. The system can self-repair and reorganize like a living organism, ensuring continuous combat effectiveness and high reliability in long-term, high-intensity combat missions.

[0163] As can be seen from the detailed description of this embodiment, the system of the present invention is not a simple stacking of modules, but a fundamental paradigm shift from "open-loop, static, single-point" to "closed-loop, dynamic, and collaborative" countermeasures through a deeply integrated technical system that includes a variety of advanced artificial intelligence and collaborative control algorithms. This comprehensively and systematically improves the intelligence level, combat effectiveness, and combat adaptability of UAV countermeasures.

[0164] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0165] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0166] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A cognitive closed-loop collaborative countermeasure system based on digital source interference and deception, characterized in that, It includes a master node and at least two slave nodes, wherein the master node and each of the slave nodes are interconnected via a wired or wireless network; The master node includes: The broadband spectrum sensing module is used to perform broadband radio monitoring of the target airspace, capture and digitize the first radio frequency signal of the target UAV in real time, and continuously collect the second radio frequency signal of the target UAV under interference after the countermeasure action is executed. The sensor module is used to perform optical monitoring of the target drone and acquire the video stream of the target drone; A cognitive processing unit is connected to the broadband spectrum sensing module and the sensor module; The cognitive processing unit is configured as follows: The first radio frequency signal is subjected to spectrum analysis to extract electromagnetic features, and the video stream is processed to extract visual features, and trajectory features are generated based on the video stream and / or the first radio frequency signal; the electromagnetic features, visual features and trajectory features are correlated and analyzed through a cross-modal fusion network to output the distribution probability of the target UAV intent category; Based on the probability distribution of the intent category, an adaptive interference and deception strategy for the target UAV is generated through a meta-reinforcement learning model; By comparing and analyzing the first radio frequency signal and the second radio frequency signal, and extracting the statistical feature changes of their channel state information, a quantitative effect assessment of the countermeasures already executed is generated. The central control unit is connected to the cognitive processing unit and each of the slave nodes; The central control unit is configured to: determine a coordinated jamming and deception scheme in collaboration with each slave node based on the countermeasure mission, the adaptive jamming and deception strategy, and the specific location of the target UAV; and generate and output control commands to each slave node based on the coordinated jamming and deception scheme. Receive the quantitative effect evaluation, and dynamically adjust the collaborative interference and deception scheme and corresponding control commands based on the quantitative effect evaluation; The slave nodes include: A multi-channel digital source module is connected to the central control unit to receive the control commands and generate and output corresponding interference signals or navigation deception signals according to the control commands. The central control unit is configured to generate and output the control commands through the following steps: Based on the countermeasures mission, the adaptive jamming and deception strategy, and the specific location of the target UAV, the central control unit defines the strategy space and utility function for each slave node. The utility function for: in, To determine the probability of a successful counterattack. The energy consumption of the slave node itself. The probability of being detected by the enemy. , , These are the corresponding weighting coefficients; Each slave node autonomously iteratively updates its policy based on the utility function using a distributed decision algorithm, which is a log-based linear learning algorithm. At any moment Choose the strategy with the following probabilities : in, This indicates that during the summation process, the nodes are traversed. strategy space Every possible strategy taken at that time Indicates the strategy of other slave nodes. Parameters for controlling the degree of exploration, For the node The strategy space; Through the iteration of the distributed decision-making algorithm, the strategies of each slave node converge to the Nash equilibrium point. The central control unit combines the strategies adopted by each slave node at the equilibrium point and integrates them into the cooperative interference and deception scheme. Based on the cooperative interference and deception scheme, it generates and outputs control commands to each slave node.

2. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 1, characterized in that, Both the master node and the slave node include: The BeiDou / GPS dual-mode timing module is used to provide the node with a nanosecond-level absolute UTC time reference. The IEEE 1588 precision clock protocol processing unit is used to run a precision clock synchronization protocol between the master node and each of the slave nodes, so that the master node and each of the slave nodes keep time synchronized. Wherein, the master node is the master clock source of the precision clock synchronization protocol, and the slave node is the slave clock of the precision clock synchronization protocol; Based on the time synchronization, the control commands output by the central control unit include the transmission time defined based on absolute UTC time; The slave node compensates for its internal processing delay and the signal propagation delay to the master node based on the transmission time, so as to control the multi-channel digital source module to accurately transmit interference signals or navigation decoy signals at the transmission time. The compensation is calculated using the following formula to obtain the start time of the interference signal or navigation decoy signal: in, Due to internal processing delay, For signal propagation delay, This refers to the launch time.

3. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 1, characterized in that, The cognitive processing unit is configured to generate the distribution probability of the target drone intent category through the following steps: The first radio frequency signal is subjected to spectrum analysis to extract electromagnetic features. The target UAV is locked from the video stream using the YOLOX target detection algorithm. The visual features are extracted using a convolutional neural network. The target UAV's flight trajectory is generated by visual tracking using the video stream and / or by TDOA direction finding using the first radio frequency signal. The trajectory features are extracted based on the flight trajectory. The electromagnetic features, visual features, and trajectory features are encoded into three independent feature vector sequences, and these three independent feature vector sequences are input into a cross-modal deep learning fusion network for association analysis, outputting the distribution probability of the target UAV's intent category, specifically: in, , , These are the query vector, key vector, and value vector, respectively, obtained by linearly transforming the feature vector sequences corresponding to the electromagnetic features, visual features, and trajectory features. is the dimension of the key vector.

4. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 1, characterized in that, The construction and use of the meta-reinforcement learning model includes the following steps: In the pre-training phase, a training task distribution containing multiple UAV communication models is constructed in a simulation environment. Meta-training is performed on all tasks based on a meta-learning algorithm to obtain initial policy model parameters with cross-task transferability. Specifically: in, These are the initial policy model parameters. Represents the first in the training task distribution One task, The learning rate within the task. Representative to Mathematical operators for finding gradients For the task loss function For parameters gradient, In the mission The adaptation strategy parameters obtained after fine-tuning; During the online adaptation phase, when encountering an unknown target drone, the cognitive processing unit defines the current scenario as a new meta-task, loads the initial strategy model parameters θ, and performs rapid gradient updates on the parameters θ through exploratory interactions with the target drone and based on the instantaneous reward signal obtained from the quantitative effect evaluation, to obtain optimized strategy parameters for the current target drone. Based on the optimized strategy parameters, an adaptive interference and deception strategy adapted to the current target drone is generated.

5. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 1, characterized in that, The cognitive processing unit is configured to generate the quantitative performance evaluation through the following steps: Extract the channel state information of the first radio frequency signal and the second radio frequency signal; Calculate the statistical feature change of the channel state information of the second radio frequency signal relative to the channel state information of the first radio frequency signal to construct an interference fingerprint feature vector. The statistical features include at least amplitude perturbation entropy and phase noise power. The interference fingerprint feature vector is input into a pre-trained interference effect classifier; the output is a quantitative evaluation result of the countermeasure effect, which is used to characterize the effectiveness level of interference or deception.

6. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 1, characterized in that, The multi-channel digital source module includes a direct digital synthesis unit, a field-programmable gate array, a high-speed digital-to-analog converter, a control unit, and a high-linearity power amplifier link; The multi-channel digital source module is configured as follows: According to the control command, the corresponding baseband I / Q waveform signal is dynamically generated in the field programmable gate array; The baseband I / Q waveform signal is converted into an analog radio frequency signal using the direct digital synthesis unit and high-speed digital-to-analog converter. The analog radio frequency signal is amplified by the high linearity power amplification link and output as the interference signal or navigation decoy signal; The multi-channel digital source module supports generating radio frequency signals with a frequency range of 100MHz to 6GHz, supports multiple modulation methods including white noise, QPSK, BPSK, and OFDM, and has at least 4 independent signal output channels.

7. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 6, characterized in that, The field-programmable gate array is configured such that its internal signal generation pipeline parameters, including carrier frequency, modulation type, symbol rate, and filter coefficients, can be dynamically reconfigured according to the control command within microseconds. Furthermore, the power control of the multi-channel digital source module adopts a combination of digital predistortion and closed-loop power detection, specifically: When generating the baseband I / Q waveform signal, the field-programmable gate array performs digital amplitude predistortion scaling according to the target power value in the control command; The output of the high linearity power amplifier link is equipped with a power detection circuit to detect the actual output power and feed it back to the control unit; The control unit is configured to perform closed-loop adjustment of the digital amplitude predistortion scaling based on the target power value and the actual output power feedback, so that the transmission power accuracy of the interference signal or navigation decoy signal is better than ±1dB.

8. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 2, characterized in that, When at least two slave nodes need to perform focused jamming on the same distant target UAV. The central control unit is also configured to: calculate the signal propagation path difference from each node to the target UAV, and convert the propagation path difference into an accurate transmission time difference or carrier initial phase difference; Generate differentiated control commands for each slave node, including the transmission time difference or the initial phase difference of the carrier. Each slave node controls the multi-channel digital source module to generate and transmit interference signals or navigation decoy signals according to the differentiated control instructions, so that the interference signals or navigation decoy signals transmitted by multiple slave nodes are spatially superimposed at the location of the target UAV.

9. The cognitive closed-loop collaborative countermeasure system based on digital source interference and deception according to claim 1, characterized in that, The system also includes state synchronization and dynamic fault tolerance mechanisms, specifically: Each of the slave nodes is configured to periodically output its own status information data to the central control unit of the master node. The status information data includes at least health status, remaining power capacity, and synchronization lock status. The central control unit is configured to monitor the status information data of each slave node in real time. When it is determined that any slave node has failed or has insufficient performance, based on the quantitative effect evaluation and the status of the remaining slave nodes, the determination process of the cooperative interference and deception scheme is re-executed, and the updated control commands are dynamically distributed to other slave nodes.

Images