Unmanned aerial vehicle signal intelligent detection and countermeasure method based on deep learning

Abstract

The application relates to the technical field of unmanned aerial vehicle signal processing, and discloses an unmanned aerial vehicle signal intelligent detection and countermeasure method based on deep learning, which collects real-time data of unmanned aerial vehicle signals through a multi-source sensor, extracts and fuses features by using a convolutional neural network, and then classifies the signals through a generative adversarial network. A signal countermeasure model is constructed, an improved particle swarm algorithm is used to optimize the countermeasure strategy, and the interference effect and energy consumption are balanced. A hierarchical countermeasure control model is established, the decision layer is globally planned, the execution layer is locally interfered, and the feedback layer is real-time evaluated and adjusted. The method has high detection precision and signal recognition accuracy, can efficiently optimize the countermeasure strategy, realizes global and local collaborative countermeasures and real-time dynamic adjustment, has wide applicability and strong expandability, can effectively cope with various unmanned aerial vehicle security threats, and guarantees public safety, privacy and important regional order.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) signal processing technology, specifically to a method for intelligent detection and countermeasures of UAV signals based on deep learning. Background Technology

[0002] In this era of rapid technological advancement, drone technology has gained widespread application in numerous fields due to its unique advantages, such as flexibility, cost-effectiveness, and ability to perform dangerous or complex tasks. In aerial photography, drones can capture stunning natural landscapes and cityscapes; in agricultural production, they can be used for farmland monitoring and pesticide spraying, improving production efficiency; and in logistics and delivery, drones are attempting to achieve "last-mile" delivery, overcoming geographical limitations. However, the widespread use of drones has also brought about a series of serious safety and management challenges.

[0003] In the realm of public safety, unauthorized drones may intrude into sensitive areas such as airports, military bases, and government agencies. Once in these areas, these drones could disrupt normal aviation operations, affect flight safety, and even be maliciously used to carry dangerous items that threaten critical facilities or personnel. Regarding privacy, the high-definition cameras carried by drones could potentially film private areas without people's knowledge, infringing on personal privacy. At large events, such as sporting events and concerts, the concentration of numerous drones could cause signal interference, leading to communication disruptions and severely impacting the normal conduct of the event.

[0004] To address these issues, drone signal detection and countermeasures technologies have emerged. Early detection primarily relied on radio frequency identification (RFID) technology, which identifies targets by recognizing specific radio frequency signals emitted by drones. However, this technology is susceptible to interference from other radio frequency signals in the environment, resulting in a high false alarm rate. Furthermore, its effectiveness is poor against some modified drones or those employing special communication protocols. Later, radar-based detection technologies appeared. While radar can accurately measure the distance, speed, and azimuth of drones, its small radar cross-section limits its detection range for small drones, and its performance is significantly affected by terrain and weather conditions, particularly in mountainous or foggy environments.

[0005] In terms of countermeasures, traditional jamming techniques mainly involve emitting high-power radio frequency interference signals to disrupt the communication link between the drone and its remote controller. However, this method lacks specificity and can not only interfere with the target drone but also affect legitimate communication equipment in the vicinity, causing communication failures. Moreover, with the development of drone technology, some drones have acquired anti-jamming capabilities, making traditional jamming methods ineffective. Additionally, there are physical capture methods, such as using net guns or drone "killer" devices to capture drones. However, these methods require close proximity, are difficult to operate, and may damage the drone or the surrounding environment.

[0006] With the rise of deep learning technology, it has achieved great success in fields such as image recognition and speech processing. Deep learning possesses powerful feature learning and pattern recognition capabilities, automatically extracting key features from large amounts of data, and exhibiting high accuracy in target recognition and classification in complex environments. In the field of drone signal detection and countermeasures, the introduction of deep learning technology has become a research hotspot. Currently, there are some deep learning-based drone detection methods, but most rely solely on data from a single sensor, failing to fully utilize the multifaceted characteristics of drones, and thus their detection accuracy needs improvement. Regarding countermeasures, although some research has attempted to optimize countermeasure strategies using machine learning algorithms, there are still shortcomings in maximizing interference effectiveness and balancing energy consumption, and an effective feedback mechanism is lacking to adjust countermeasures in real time. Summary of the Invention

[0007] The purpose of this invention is to provide a deep learning-based intelligent detection and countermeasure method for drone signals, in order to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a deep learning-based intelligent detection and countermeasure method for drone signals, the method comprising:

[0009] Real-time data of UAV signals are collected through a multi-source sensor, which includes a radio frequency sensor, an infrared sensor, an acoustic sensor, and an optical sensor.

[0010] The real-time data is processed by feature extraction and fusion based on a convolutional neural network to obtain signal feature data.

[0011] The signal feature data is input into a pre-trained generative adversarial network model, which employs a dual discriminator structure to identify and classify signals based on signal feature distribution, and generate signal identification results.

[0012] A signal countermeasure model is constructed based on the signal recognition results. The signal countermeasure model aims to maximize the signal interference effect and optimizes the countermeasure strategy using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm introduces adaptive inertial weights and dynamic learning factors.

[0013] The optimal countermeasure strategy is output based on the aforementioned signal countermeasure model.

[0014] A hierarchical countermeasure control model is established based on the optimal countermeasure strategy. The hierarchical countermeasure control model includes a decision layer, an execution layer, and a feedback layer. The decision layer plans a global countermeasure strategy based on the signal recognition results, the execution layer performs local signal interference based on the optimal countermeasure strategy, and the feedback layer realizes real-time evaluation and adjustment of the countermeasure effect based on an adaptive filtering algorithm.

[0015] The hierarchical countermeasure control model outputs countermeasure control commands to achieve intelligent detection and countermeasure against UAV signals.

[0016] Preferably, the process of constructing the generative adversarial network model includes:

[0017] Acquire real-time signal data, which includes signal frequency, signal strength, signal modulation method, and signal source location;

[0018] A signal feature space is constructed based on the real-time signal data, and a classification space is constructed based on the distinguishability of the signal features.

[0019] A multi-objective loss function is constructed based on the signal feature space and the classification space. The multi-objective loss function includes a classification loss term, a feature matching loss term, and an adversarial loss term.

[0020] A dual discriminator structure is constructed, which includes a main discriminator and an auxiliary discriminator. Both the main discriminator and the auxiliary discriminator include an input layer, three hidden layers, and an output layer.

[0021] The generative adversarial network model is trained using a batch training method;

[0022] The signal identification results are output based on the trained generative adversarial network model. The signal identification results include signal category, signal source location, and signal threat level.

[0023] Preferably, the multi-objective loss function is specifically composed of:

[0024] The classification loss term is calculated using the cross-entropy between the signal features and the classification label;

[0025] The feature matching loss term is calculated using the Euclidean distance between the signal features and the generated features;

[0026] The adversarial loss term is calculated by the KL divergence between the probability distribution output by the discriminator and the true distribution.

[0027] Preferably, the dual discriminator structure specifically includes:

[0028] The dimension of the input layer is the same as the dimension of the signal feature space;

[0029] The hidden layer uses the LeakyReLU activation function;

[0030] The output layer has the same dimension as the classification space and uses the Softmax activation function.

[0031] A joint loss function is constructed based on the main discriminator and the auxiliary discriminator, the joint loss function including the main discriminator loss term and the auxiliary discriminator loss term.

[0032] Preferably, the process of constructing the signal countermeasure model includes:

[0033] Construct a multi-objective function for signal countermeasures, the multi-objective function including an interference effect objective function and an energy consumption objective function;

[0034] The countermeasure strategy constraints are constructed based on the multi-objective function, and the countermeasure strategy constraints include power constraints, frequency constraints, time constraints, and distance constraints.

[0035] The countermeasure strategy is encoded using a particle swarm optimization algorithm, with each particle containing information on interference power, interference frequency, and interference time.

[0036] An adaptive inertia weight mechanism and a dynamic learning factor are introduced for iterative optimization;

[0037] The optimal countermeasure strategy is selected from the Pareto front solution set that satisfies the trade-off between interference effect and energy consumption.

[0038] Preferably, the specific composition of the multi-objective function includes:

[0039] The objective function for the interference effect is calculated by multiplying the signal strength attenuation rate by the interference time.

[0040] The energy consumption objective function is calculated by multiplying the interference power by the interference time;

[0041] The particle fitness is obtained by weighting the interference effect and energy consumption.

[0042] Preferably, the optimization process of the improved particle swarm optimization algorithm includes:

[0043] The adaptive inertia weight mechanism employs a time-varying inertia weight factor, which exhibits a linear decay as the number of iterations increases.

[0044] The dynamic learning factor increases linearly with the number of iterations.

[0045] Based on the adaptive inertia weight mechanism and the dynamic learning factor, the interference effect and energy consumption of the countermeasure strategy generated in each iteration are evaluated, and the non-dominated solution is added to the Pareto front solution set.

[0046] Preferably, the decision-making layer's global countermeasure strategy planning based on the signal recognition results includes:

[0047] The countermeasure strategy is described by a parameterized curve, and the countermeasure strategy is expressed as a function of the countermeasure strategy parameters. The values ​​of the countermeasure strategy parameters range from 0 to 1. The countermeasure strategy includes interference power components, interference frequency components, and interference time components.

[0048] The countermeasure strategy is described based on cubic spline curves. The parameter values ​​of the countermeasure strategy are calculated by summing the products of the combination coefficients and the spline basis functions. The spline basis functions are calculated by power functions of the combination number and the countermeasure strategy parameters.

[0049] Constructing countermeasure strategy constraints, which include power constraints, frequency constraints, time constraints, and distance constraints. The power constraints are used to limit the power range of the countermeasure strategy, the frequency constraints are used to limit the frequency range of the countermeasure strategy, the time constraints are used to limit the time range of the countermeasure strategy, and the distance constraints are used to limit the distance range of the countermeasure strategy.

[0050] A global multi-objective optimization function is constructed, which includes a disturbance effect term, an energy consumption term, a disturbance time term, and a disturbance distance term. The terms in the global multi-objective optimization function are weighted and combined using weighting coefficients.

[0051] The countermeasure strategy interval is discretized into multiple strategy segments, and the global multi-objective optimization function is discretized to construct a global discretized objective function. The global discretized objective function includes the interference effect of the strategy segment, the energy consumption of the strategy segment, the time of the strategy segment, and the distance of the strategy segment.

[0052] The global discretized objective function is iteratively optimized using a sequential quadratic programming method. By constructing a quadratic approximation of the global multi-objective optimization function and linear constraints, the parameter increments of the control points are obtained by solving the quadratic programming problem. The parameter values ​​of the control points are then updated based on these parameter increments.

[0053] The optimized countermeasure strategy is smoothed by fifth-order spline interpolation. By maintaining the continuity of the parameter derivative, power derivative, and frequency derivative at the interpolation endpoints, a smooth and continuous countermeasure strategy is generated. Based on the smooth and continuous countermeasure strategy, an interference parameter sequence and corresponding interference control commands are generated.

[0054] Preferably, the execution layer performs local signal interference based on the optimal countermeasure strategy, including:

[0055] A local interference window is constructed based on the current location, current power, and current frequency of the interference device. The size of the local interference window is adaptively adjusted by a power adaptive coefficient, and a positive correlation is established between the size of the local interference window and the magnitude of the current power.

[0056] A local signal environment model is constructed using multi-source sensor data. The sensor data is transformed to obtain signal data in the local coordinate system. The signal intensity distribution map is updated based on the signal data. The intensity value of each signal point is calculated using a logarithmic probability update method.

[0057] The Kalman filter algorithm is used to track the dynamic signal source. The state vector of the signal source is predicted by the state prediction equation. The predicted state is updated based on the measurement data to obtain the precise position and motion information of the signal source.

[0058] An interference optimization model is constructed, and the kinematic equation of the interference device is used as the state equation. The state equation includes position coordinates, power and frequency. State constraints and dynamic constraints are constructed. The state constraints are used to limit the range of values ​​for position coordinates and power. The dynamic constraints are used to limit the range of values ​​for frequency change rate, power change rate and time change rate.

[0059] A multi-objective cost function is constructed, which includes a reference interference effect term, a signal source avoidance term, an interference smoothing term, and an energy consumption term. The terms in the multi-objective cost function are weighted and combined using weighting coefficients.

[0060] The gradient projection method is used to optimize the multi-objective cost function, calculate the gradient of the cost function with respect to the control quantity, project the updated control quantity to the feasible region, and determine the optimal step size through backtracking search.

[0061] Preferably, the feedback layer, based on an adaptive filtering algorithm, implements real-time evaluation and adjustment of the countermeasure effect, including:

[0062] An interference effect evaluation model is established, which includes a signal strength attenuation equation and an interference time equation. The signal strength attenuation equation includes an interference power term, an interference frequency term, and an interference time term. The interference time equation includes an interference power term and an interference frequency term.

[0063] The interference effect evaluation model is constructed as a state-space expression. The state vector of the state-space expression includes the signal strength attenuation rate, interference time, and interference distance. The control vector of the state-space expression includes the interference power and interference frequency.

[0064] The state-space expression is linearized, and the Jacobian matrix of the system state equation with respect to the state vector and control vector is calculated to construct a linearized prediction model.

[0065] A cost function for predicting interference effects is constructed. The cost function includes an interference effect error term, a control quantity penalty term, and a control increment penalty term. The penalty terms are weighted and combined using a weight matrix.

[0066] State constraints are constructed, including signal strength attenuation rate constraints and interference time constraints. Control constraints are constructed, including interference power constraints and interference frequency constraints. Control increment constraints are constructed, including interference power increment constraints and interference frequency increment constraints.

[0067] The prediction cost function is transformed into the standard form of a quadratic programming problem, and the quadratic form matrix and the coefficients of the linear terms are calculated to construct the inequality constraint matrix and the equality constraint matrix.

[0068] The quadratic programming problem is solved using the interior point method. Based on the optimized solution, the control variables are mapped. The total interference power is allocated to each interference device through the power allocation matrix, and the interference frequency is allocated to each interference device through the frequency allocation matrix.

[0069] The output of the interference device is limited, the interference power is limited according to the power characteristics, the interference frequency is limited according to the frequency characteristics, and the final interference control command is generated.

[0070] Compared with the prior art, the beneficial effects of the present invention are:

[0071] This invention employs multi-source sensor data acquisition, combined with the powerful feature extraction and fusion capabilities of convolutional neural networks, to comprehensively and accurately capture the signal characteristics of drones. Radio frequency (RF) sensors acquire the RF signal characteristics of the drone, while infrared, acoustic, and optical sensors supplement information from different dimensions. After processing by the convolutional neural network, the resulting signal feature data is richer and more accurate. This process effectively overcomes the shortcomings of traditional single-sensor detection methods, such as susceptibility to environmental interference and incomplete information acquisition, significantly improving the accuracy and reliability of drone signal detection and greatly reducing the probability of false alarms and missed alarms.

[0072] The generative adversarial network (GAN) model employs a dual discriminator structure and is trained using a multi-objective loss function. This allows for in-depth analysis of signal feature distribution, enabling accurate identification of signal categories, determination of signal source locations, and assessment of signal threat levels. The classification loss term, feature matching loss term, and adversarial loss term in the multi-objective loss function optimize model performance from different perspectives, ensuring stable and accurate operation even in complex signal environments. Compared to traditional identification methods, this model demonstrates stronger identification capabilities for novel UAV signals or camouflaged / jammed signals, providing a reliable basis for developing targeted countermeasures.

[0073] The signal countermeasure model aims to maximize signal interference effectiveness and optimizes the countermeasure strategy using an improved particle swarm optimization (PSO) algorithm. The improved PSO algorithm, incorporating adaptive inertia weights and dynamic learning factors, better balances global and local search capabilities during the optimal solution search process, avoiding getting trapped in local optima. By constructing multi-objective functions (including interference effectiveness and energy consumption objective functions) and corresponding constraints, and comprehensively considering factors such as interference effectiveness and energy consumption, the optimal countermeasure strategy is selected from the Pareto front solution set. This effectively interferes with UAV signals while reducing energy consumption and improving countermeasure efficiency. Compared to traditional countermeasure methods, this optimization strategy improves resource utilization and reduces unnecessary energy waste while ensuring countermeasure effectiveness.

[0074] The hierarchical countermeasure control model's decision layer plans a global countermeasure strategy based on signal recognition results. It uses parametric curves and cubic spline curves to describe the countermeasure strategy and constructs a global multi-objective optimization function. Through iterative optimization using a sequential quadratic programming method, it generates a smooth and continuous sequence of countermeasure strategies and interference parameters. This process fully considers various constraints, achieving a globally rational allocation of interference resources and ensuring the scientific and global nature of the countermeasure strategy. The execution layer performs local signal interference based on the optimal countermeasure strategy. Through operations such as constructing local interference windows, tracking dynamic signal sources, and establishing interference optimization models, it achieves accurate perception and interference of the local signal environment. The collaborative work between the decision layer and the execution layer enables the countermeasure process to grasp the overall situation macroscopically while flexibly adjusting to local conditions, significantly improving the targeting and effectiveness of the countermeasures.

[0075] The feedback layer, based on an adaptive filtering algorithm, achieves real-time evaluation and adjustment of the countermeasures effect through a series of operations, including establishing an interference effect evaluation model, constructing a state-space expression, and predicting a cost function. Based on the evaluation results, the feedback layer can promptly adjust control variables such as interference power and frequency, and distribute the adjusted parameters to each interference device through power and frequency allocation matrices, ensuring the countermeasures effect remains optimal. This real-time feedback mechanism makes the countermeasures system adaptive, enabling it to cope with complex and ever-changing UAV signal environments and ensuring continuous and stable countermeasures under different conditions, greatly improving the performance and reliability of the entire countermeasures system.

[0076] The detection and countermeasure method of this invention does not rely on specific drone models or communication protocols, and can adapt to various types of drone signals, showing good application prospects in different scenarios. It can be effectively used in complex areas such as urban environments, mountainous regions, airport perimeters, and military bases. Furthermore, this method is based on deep learning and modular design, making it easy to expand and upgrade. With the continuous development of drone technology, the system performance can be further improved by updating training data and optimizing the model structure to address emerging drone threats, demonstrating strong sustainable development capabilities. Attached Figure Description

[0077] Figure 1 This is a schematic diagram illustrating the working principle of the deep learning-based intelligent detection and countermeasure method for drone signals described in this invention.

[0078] Figure 2 A flowchart for constructing and training a generative adversarial network model;

[0079] Figure 3 A flowchart for building and optimizing strategies for signal countermeasure models. Detailed Implementation

[0080] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] Please see Figure 1-3 This invention provides a technical solution: a method for intelligent detection and countermeasure of drone signals based on deep learning, the method comprising:

[0082] Data Acquisition: Real-time data from drone signals is acquired using multi-source sensors, including radio frequency (RF) sensors, infrared sensors, acoustic sensors, and optical sensors. RF sensors receive the RF signals emitted by the drone, obtaining information such as signal frequency and intensity. Infrared sensors detect the infrared radiation emitted by the drone, aiding in its location. Acoustic sensors capture the sound signals generated during drone flight, providing a basis for determining the drone's position and type. Optical sensors acquire images, providing a visual representation of the drone's appearance and location. These sensors collect data from different perspectives, providing comprehensive information for subsequent processing.

[0083] Feature Extraction and Fusion: Convolutional neural networks (CNNs) are used to extract and fuse features from the acquired real-time data. CNNs possess powerful feature extraction capabilities; through structures such as convolutional layers, pooling layers, and fully connected layers, they automatically extract key signal features from multi-source sensor data and fuse these features to form more representative signal feature data.

[0084] Signal recognition and classification: The signal feature data obtained through feature extraction and fusion is input into a pre-trained generative adversarial network (GAN) model. This GAN model employs a dual discriminator structure to recognize and classify signals based on signal feature distribution, thereby generating signal recognition results. This process can accurately determine information such as signal type, signal source location, and signal threat level.

[0085] Constructing a Signal Countermeasure Model: A signal countermeasure model is constructed based on the signal identification results. This model aims to maximize the signal interference effect and optimizes the countermeasure strategy using an improved particle swarm optimization (PSO) algorithm. The improved PSO algorithm incorporates adaptive inertia weights and a dynamic learning factor, enabling it to better balance global and local search capabilities during the optimal solution search process, thereby improving optimization efficiency.

[0086] Output the optimal countermeasure strategy: Based on the constructed signal countermeasure model, output the optimal countermeasure strategy. This strategy comprehensively considers factors such as interference effect and energy consumption, and is the optimal choice under certain constraints.

[0087] A hierarchical countermeasure control model is established: Based on the optimal countermeasure strategy, a hierarchical countermeasure control model is established, which includes a decision-making layer, an execution layer, and a feedback layer. The decision-making layer plans the global countermeasure strategy based on the signal recognition results; the execution layer implements local signal interference according to the optimal countermeasure strategy; and the feedback layer uses an adaptive filtering algorithm to evaluate and adjust the countermeasure effect in real time, forming a closed-loop control system to ensure the effectiveness and stability of the countermeasures.

[0088] Intelligent detection and countermeasures: By outputting countermeasure control commands through a hierarchical countermeasure control model, corresponding jamming devices are driven to interfere with the UAV signal, thereby achieving intelligent detection and countermeasures of the UAV signal.

[0089] The present invention will be further described below with reference to Examples 1 to 5:

[0090] Example 1:

[0091] This embodiment mainly illustrates in detail the construction process of the generative adversarial network model, which is as follows:

[0092] ① Acquire real-time signal data: Acquire real-time signal data from various monitoring devices, covering signal frequencies ( ), signal strength ( ), signal modulation method ( ) and the location of the signal source ( For example, in practical monitoring scenarios, signal frequency and intensity are acquired through radio frequency monitoring equipment. The signal modulation method is obtained from communication protocol parsing, and the location of the signal source is determined using positioning technology.

[0093] ② Constructing the Signal Feature Space and Classification Space: A signal feature space is constructed based on the acquired real-time signal data. This space is an abstract representation of various signal characteristics, incorporating quantified and encoded information such as signal frequency, intensity, modulation scheme, and signal source location. Simultaneously, a classification space is constructed based on the distinguishability of the signal features. This space defines the possible classification results of the signal. For example, signals can be categorized into different classes based on the drone's model and purpose, with each class having a clear definition within the classification space.

[0094] ③ Construct a multi-objective loss function: Construct a multi-objective loss function, which includes a classification loss term ( ), feature matching loss term ( ) and counter-loss items ( ).

[0095] The classification loss term is calculated using the cross-entropy between the signal features and the classification label, as shown in the formula: ,in These are the actual category labels (with values ​​of 0 or 1). It is the probability that the model predicts belongs to that category. This represents the total number of categories.

[0096] The feature matching loss term is calculated using the Euclidean distance between the signal features and the generated features, assuming the true signal feature vector is... The generated feature vector is ,but , is the dimension of the feature vector.

[0097] The adversarial loss term is calculated using the KL divergence between the probability distribution output by the discriminator and the true distribution, as shown in the formula: , It is the probability of the true distribution. It is the probability of generating the distribution. is the dimension of the distribution.

[0098] ④ Construct a dual discriminator structure: Construct a dual discriminator structure, including a main discriminator and an auxiliary discriminator. Both the main discriminator and the auxiliary discriminator consist of an input layer, three hidden layers, and an output layer.

[0099] The dimension of the input layer is the same as the dimension of the signal feature space, ensuring that complete signal feature data can be received.

[0100] The hidden layer uses the LeakyRelU activation function, whose formula is:

[0101]

[0102] generally Choose a smaller value, such as 0.01, to avoid... The problem of neurons not being activated at all.

[0103] The output layer has the same dimension as the classification space and uses the Softmax activation function. The Softmax function formula is:

[0104]

[0105] These are the elements in the input vector. The output is a vector dimension, and the Softmax function is used to transform it into a probability distribution, which facilitates classification.

[0106] A joint loss function is constructed based on the main discriminator and the auxiliary discriminator. The joint loss function includes the loss term of the main discriminator ( ) and the loss term of the auxiliary discriminator ), joint loss function .

[0107] ⑤ Training the Generative Adversarial Network (GAN) Model: The GAN model is trained using a batch training method. During training, the parameters of the generator and discriminator are continuously adjusted so that the signal features generated by the generator can confuse the discriminator as much as possible, while the discriminator can accurately distinguish between real and generated signal features. After multiple rounds of training, the model gradually converges, and its performance continuously improves.

[0108] ⑥ Output signal recognition results: Based on the trained generative adversarial network model, new signal feature data is input, and the model outputs signal recognition results, including signal category, signal source location, and signal threat level. For example, the model determines that a signal belongs to a specific type of drone, the signal source is located at a specific coordinate location, and its threat level is assessed as high, medium, or low based on the signal characteristics.

[0109] Example 2:

[0110] This embodiment aims to explain in detail the construction process of the signal countermeasure model. The specific steps are as follows:

[0111] ① Constructing a multi-objective function: Constructing a multi-objective function for signal countermeasures, including an objective function for interference effects ( ) and energy consumption objective function ( The objective function for interference effect is expressed as the signal strength attenuation rate (). ) and interference time ( The product of ) is calculated using the following formula: For example, suppose that at a certain moment, interference with the drone signal causes a signal strength attenuation rate of 0.5, and the interference duration is 10 seconds. Then, the objective function value of the interference effect at that moment is... .

[0112] The energy consumption objective function is obtained through interference power ( ) and interference time ( The product of ) is calculated using the following formula: If the interference power is 100 watts and the interference time is 10 seconds, then the energy consumption objective function value is... joule.

[0113] ② Constructing countermeasure strategy constraints: Construct countermeasure strategy constraints based on multi-objective functions, including power constraints, frequency constraints, time constraints, and distance constraints.

[0114] Power constraints are used to limit the power range of countermeasure strategies. Let the maximum power be... Minimum power is Then the power constraint condition is For example, if the maximum power of the interference device is 500 watts and the minimum power is 50 watts, then the interference power must be within this range.

[0115] Frequency constraints are used to limit the frequency range of countermeasures; let the maximum frequency be... The minimum frequency is The frequency constraint is Assuming the jamming device can operate in the frequency range of 1GHz-10GHz, then the jamming frequency can only be selected within this range.

[0116] Time constraints are used to limit the time range of the countermeasure strategy. Let the maximum interference time be... The shortest interference time is The time constraint is For example, the shortest duration of a single interference operation must be no less than 1 second, and the longest duration must not exceed 60 seconds.

[0117] Distance constraints are used to limit the range of countermeasures. Let the maximum interference distance be... The minimum interference distance is The distance constraint is If the effective interference range of the jamming device is 100 meters to 1000 meters, then the interference distance must be within this range.

[0118] ③ Particle Swarm Optimization (PSO) Encoding: The countermeasure strategy is encoded using the Particle Swarm Optimization (PSO) algorithm, with each particle containing interference power (PSO). ), interference frequency ( ) and interference time ( Information. These parameters are used as the positions of particles in the solution space. The particle swarm optimization algorithm is used to continuously adjust the particle positions to find the optimal countermeasure strategy.

[0119] ④ Iterative optimization: Adaptive inertia weight mechanism and dynamic learning factor are introduced for iterative optimization.

[0120] The adaptive inertia weighting mechanism uses a time-varying inertia weighting factor. The time-varying inertia weighting factor varies with the number of iterations. The increase exhibits a linear decay, as shown by the formula: ,in It is the initial inertia weight. It is the final inertia weight. This is the maximum number of iterations. For example, , , When the number of iterations hour,

[0121] The dynamic learning factor increases linearly with the number of iterations. Let the learning factor... and , , , , , , These are the minimum and maximum values ​​of the learning factor, respectively. For example... , , , , ,when hour, , .

[0122] Based on an adaptive inertia weighting mechanism and a dynamic learning factor, the interference effect and energy consumption of the countermeasure strategy generated in each iteration are evaluated, and non-dominated solutions are added to the Pareto front solution set. A non-dominated solution is a solution in multi-objective optimization where no other solution is superior to it in all objectives. Through continuous iteration, the Pareto front solution set gradually approaches the optimal solution set.

[0123] ⑤ Select the optimal solution: Select the optimal solution from the Pareto front solution set that satisfies the trade-off between interference effect and energy consumption as the optimal countermeasure strategy. For example, in the Pareto front solution set, depending on actual needs, more emphasis may be placed on the interference effect. In this case, the solution with relatively good interference effect and acceptable energy consumption can be selected as the optimal countermeasure strategy; if energy consumption is more sensitive, the solution with lower energy consumption and interference effect that meets the basic requirements can be selected.

[0124] Example 3:

[0125] This embodiment details the process by which the decision-making layer plans a global countermeasure strategy based on signal recognition results, as follows:

[0126] ① Parametric Curve Description of Countermeasure Strategy: The countermeasure strategy is described using parametric curves, which represent the countermeasure strategy as countermeasure strategy parameters ( The parameters of the countermeasure strategy function range from 0 to 1. The countermeasure strategy includes the interference power component. Interference frequency components and interference time components ,Right now , , By adjusting the countermeasure strategy parameters, different combinations of countermeasure strategies can be obtained.

[0127] ② Describing Countermeasures Based on Cubic Spline Curves: Countermeasures are described using cubic spline curves. Cubic spline curves are described through combinations of coefficients (…). ) and spline basis functions ( The summation of the products of the two sides yields the parameter values ​​for the countermeasure strategy, as shown in the formula: Spline basis functions are obtained through combinations ( The parameters of the countermeasure strategy are calculated using a power function, for example... ,in Let be the order of the spline curve, in a cubic spline curve. By adjusting the combination coefficients, the shape of the spline curve can be changed, thereby obtaining different countermeasure strategy parameter values.

[0128] ③ Constructing countermeasure strategy constraints: Constructing countermeasure strategy constraints, including power constraints, frequency constraints, time constraints, and distance constraints.

[0129] Power constraints are used to limit the power range of countermeasure strategies. Let the maximum power be... Minimum power is ,but .

[0130] Frequency constraints are used to limit the frequency range of countermeasures; let the maximum frequency be... The minimum frequency is ,but .

[0131] Time constraints are used to limit the time range of the countermeasure strategy. Let the maximum interference time be... The shortest interference time is ,but .

[0132] Distance constraints are used to limit the range of countermeasures. Let the maximum interference distance be... The minimum interference distance is ,but These constraints ensure the feasibility of countermeasures in practical applications.

[0133] ④ Construct a global multi-objective optimization function: Construct a global multi-objective optimization function, including the interference effect term. Energy consumption items Interference time item and interference distance term Through weighting coefficients , , , The formula for weighted combination of the items is: The interference effect term can be calculated based on the interference effect objective function, the energy consumption term based on the energy consumption objective function, and the interference time and interference distance terms quantified based on the actual interference time and distance. The weighting coefficients are adjusted according to actual needs to balance the importance of different objectives.

[0134] ⑤ Discretization: The countermeasure strategy interval is discretized into multiple strategy segments. The global multi-objective optimization function is discretized to construct a global discretized objective function. The global discretized objective function includes the interference effect of the strategy segments (…). ), Strategy segment energy consumption ( ), strategy period ( ) and strategy segment distance ( For example, the countermeasure strategy interval [0, 1] can be discretized into... There are several strategy segments, each corresponding to different countermeasure strategy parameter values. The various indicators of each strategy segment are calculated to form a global discretized objective function.

[0135] ⑥ Iterative Optimization: A sequential quadratic programming method is used to iteratively optimize the globally discretized objective function. By constructing a quadratic approximation of the global multi-objective optimization function and applying linear constraints, the parameter increments of the control points are obtained by solving the quadratic programming problem. ), update the parameter values ​​of control points based on parameter increments During the iteration process, the parameter values ​​are continuously adjusted, causing the global discretization objective function to gradually optimize and approach the optimal solution.

[0136] ⑦ Smoothing and Command Generation: The optimized countermeasure strategy is smoothed using quintic spline interpolation. By maintaining the continuity of the parameter derivatives, power derivatives, and frequency derivatives at the interpolation endpoints, a smooth and continuous countermeasure strategy is generated. Based on this smooth and continuous countermeasure strategy, an interference parameter sequence and corresponding interference control commands are generated. Quintic spline interpolation smoothing makes the countermeasure strategy smoother, avoids abrupt changes, and improves the stability and reliability of the interference effect. The generated interference parameter sequence and control commands are then passed to the execution layer for actual signal interference operations.

[0137] Example 4:

[0138] This embodiment aims to provide a detailed explanation of the specific operational steps for performing local signal interference at the execution layer based on the optimal countermeasure strategy. The specific content is as follows:

[0139] Construct a local interference window: based on the current location of the interference device. Current power and current frequency A local interference window is constructed based on this. An adaptive power coefficient is introduced. This coefficient is used to adaptively adjust the size of the local interference window, and to adjust the size of the local interference window. With current power The magnitudes establish a positive correlation, and its mathematical expression is: For example, when the interference device's current power... Power adaptive coefficient At that time, the size of the local interference window (The unit of measurement here can be set according to the actual situation, such as area unit square meters or volume unit cubic meters, etc., assuming it is area unit square meters). In this way, as the interference power changes, the coverage area of ​​the local interference window can be dynamically adjusted, improving the targeting and effectiveness of the interference.

[0140] Constructing a Local Signal Environment Model: A local signal environment model is constructed using multi-source sensor data. First, the data acquired by the sensors is transformed to a local coordinate system, yielding signal data in that system. Based on this signal data, the signal intensity distribution map is updated, and the intensity value of each signal point is calculated using a log-probability update method. Assume the original signal intensity of a signal point in the local coordinate system is... The logarithmic probability update formula is: The updated signal strength value is obtained after calculation using this formula. By continuously updating the signal strength distribution map, changes in the local signal environment can be monitored in real time, providing an accurate basis for subsequent interference operations.

[0141] Tracking dynamic signal sources: The Kalman filter algorithm is used to track dynamic signal sources. The Kalman filter algorithm predicts the state vector of the signal source through a state prediction equation. ,in Is Time based The state vector predicted from the time-series information. It is the state transition matrix. yes The estimated state vector at time step [time]. It is a control input matrix. yes Timing control input. Then based on measurement data. The predicted state is updated using the following formula: ,in yes The updated state vector at each time step. It is Kalman gain. It is the observation matrix. Through this series of operations, the precise location and motion information of the signal source can be obtained, thereby enabling more accurate interference.

[0142] Constructing an interference optimization model: Using the kinematic equations of the interference device as the state equations, the state equations include position coordinates. ,power and frequency , can be represented as ,in , , It refers to the velocity of the interfering device along the three coordinate axes. It is the rate of change of power. It is the rate of change of frequency. Simultaneously, state constraints and dynamic constraints are constructed. State constraints are used to limit the range of values ​​for the position coordinates and the power, for example... , , , Dynamic constraints are used to limit the range of values ​​for the rate of change of frequency, the rate of change of power, and the rate of change of time, such as... , , These constraints ensure the safety and feasibility of the interference equipment during actual operation.

[0143] ⑤ Construct a multi-objective cost function: Construct a multi-objective cost function, which includes a reference interference effect term. Signal source avoidance item Interference smoothing term and energy consumption items Through weighting coefficients , , , The weighted combination of terms in the multi-objective cost function is expressed as follows: Reference interference effect item It can be calculated based on the objective function of the interference effect to measure the degree of interference's impact on the UAV signal; signal source avoidance term. To prevent interfering devices from getting too close to important signal sources and ensure the transmission of other normal signals, the interference smoothing term can be quantified based on factors such as the distance between the interfering device and the important signal source. Used to smooth out disturbances and reduce abrupt changes, it can be calculated using factors such as the rate of change of disturbance parameters; energy consumption item. Calculated based on the energy consumption objective function, used to assess energy consumption during disturbance processes. Weighting coefficients. , , , Adjustments should be made based on actual needs to balance the importance of different objectives.

[0144] ⑥ Optimization Solution: The gradient projection method is used to optimize the multi-objective cost function. First, the gradient of the cost function with respect to control variables (such as position coordinates, power, frequency, etc.) is calculated. The updated control input is then projected onto the feasible region to ensure that the control input satisfies both state and dynamic constraints. During the projection process, the optimal step size is determined through backtracking search. This ensures that the cost function decreases towards the optimal direction in each iteration. Specifically, assume the current control quantity is... The updated control quantity is Find the appropriate one by searching the backline. , making The minimum value must be obtained while satisfying the constraints. After multiple iterations, the optimal control quantity is finally obtained, achieving effective interference with local signals.

[0145] Example 5:

[0146] This embodiment mainly describes in detail the real-time evaluation and adjustment process of the countermeasure effect based on the adaptive filtering algorithm in the feedback layer. The specific steps are as follows:

[0147] Establish an interference effect evaluation model: This model includes a signal strength attenuation equation and an interference time equation. The signal strength attenuation equation contains an interference power term. Interference frequency item and interference time item Assume the signal strength attenuation rate is Its expression is ,in , , , The coefficients can be determined through experiments or data analysis. The interference time equation includes an interference power term. Interference frequency term Assuming the duration of the interference is The equation is , , , These are the corresponding coefficients. These equations are used to quantify the relationship between the interference effect and the interference parameters, providing a basis for subsequent evaluation.

[0148] Constructing a state-space expression: The interference effect evaluation model is constructed as a state-space expression, and the state vector of the state-space expression is... ,in Interference distance; control vector The state-space expression can be represented as: ,in It is the state transition matrix. It is a control input matrix. It is the output matrix. This form allows the interference effect evaluation model to be incorporated into the system's state-space framework, facilitating analysis and control.

[0149] Linearization: The state-space expression is linearized by calculating the Jacobian matrix of the system state equations with respect to the state vectors and control vectors. Let the state equations be... Then regarding the state vector Jacobian matrix Regarding control vectors Jacobian matrix Based on these Jacobian matrices, a linearized prediction model is constructed, and the linearized state equation is: ,in and It is the linearized coefficient matrix obtained from the Jacobian matrix. and These are the linearized state and control vectors. Linearization simplifies system analysis, improves computational efficiency, and facilitates subsequent optimization solutions.

[0150] Constructing the interference effect prediction cost function: Constructing the interference effect prediction cost function, which includes the interference effect error term. Control quantity penalty items and control of incremental penalty items Through the weight matrix , , The weighted combination of each penalty item is expressed as follows: Interference effect error term Used to measure the difference between the actual interference effect and the expected interference effect; control quantity penalty term. Used to limit the size of the control quantity to avoid excessive interference; control increment penalty term. Used to suppress drastic changes in the control quantity, making the disturbance process smoother. Weight matrix , , Adjustments should be made based on actual needs to balance the importance of different penalties.

[0151] Constructing constraints: Constructing state constraints, including signal strength attenuation rate constraints. and interference time constraints Construct control constraints, including interference power constraints. Interference frequency constraints Construct control increment constraints, including interference power increment constraints. With interference frequency increment constraints These constraints ensure that the system operates within a safe and efficient range, preventing abnormal situations from occurring.

[0152] Transform the problem into a quadratic programming problem and solve it: Transform the prediction cost function into the standard form of a quadratic programming problem and calculate the quadratic form matrix. coefficient of the first term Construct the inequality constraint matrix With equality constraint matrix The standard form of a quadratic programming problem is: , , ,in These are optimization variables (in this case, control vectors). (and its increment, etc.). The interior-point method is used to solve this quadratic programming problem. The interior-point method avoids complex calculations at the boundary by finding the optimal solution within the feasible region, exhibiting good convergence and stability. Control variables are mapped based on the optimized solution results, using the power allocation matrix. The total interference power is distributed to each interference device via a frequency allocation matrix. The interference frequencies are allocated to each interference device to achieve a reasonable allocation of interference resources.

[0153] Limiting and Generation Commands: Limit the output of the interfering device by limiting the interference power based on its power characteristics. For example, when the calculated interference power... Exceeding the maximum power limit At that time, take When below the minimum power limit At that time, take Based on frequency characteristics, interference frequencies are limited; similarly, when calculating frequencies... When the frequency exceeds the range, corresponding adjustments are made. After amplitude limiting, the final interference control command is generated and fed back to the execution layer to realize real-time adjustment and optimization of the countermeasure effect, ensuring the stable and reliable performance of the entire countermeasure system.

[0154] It should be noted that, in this document, relational terms such as "first" and "second" are used only 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 process, method, article, or apparatus.

[0155] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent detection and countermeasure of UAV signals based on deep learning, characterized in that, The method includes: Real-time data of UAV signals are collected through a multi-source sensor, which includes a radio frequency sensor, an infrared sensor, an acoustic sensor, and an optical sensor. The real-time data is processed by feature extraction and fusion based on a convolutional neural network to obtain signal feature data. The signal feature data is input into a pre-trained generative adversarial network model, which employs a dual discriminator structure to identify and classify signals based on signal feature distribution, and generate signal identification results. A signal countermeasure model is constructed based on the signal recognition results. The signal countermeasure model aims to maximize the signal interference effect and optimizes the countermeasure strategy using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm introduces adaptive inertial weights and dynamic learning factors. The optimal countermeasure strategy is output based on the aforementioned signal countermeasure model. A hierarchical countermeasure control model is established based on the optimal countermeasure strategy. The hierarchical countermeasure control model includes a decision layer, an execution layer, and a feedback layer. The decision layer plans a global countermeasure strategy based on the signal recognition results, the execution layer performs local signal interference based on the optimal countermeasure strategy, and the feedback layer realizes real-time evaluation and adjustment of the countermeasure effect based on an adaptive filtering algorithm. The hierarchical countermeasure control model outputs countermeasure control commands to achieve intelligent detection and countermeasure against UAV signals. The decision-making level plans a global countermeasure strategy based on the signal recognition results, including: The countermeasure strategy is described by a parameterized curve, and the countermeasure strategy is expressed as a function of the countermeasure strategy parameters. The values ​​of the countermeasure strategy parameters range from 0 to 1. The countermeasure strategy includes interference power components, interference frequency components, and interference time components. The countermeasure strategy is described based on cubic spline curves. The parameter values ​​of the countermeasure strategy are calculated by summing the products of the combination coefficients and the spline basis functions. The spline basis functions are calculated by power functions of the combination number and the countermeasure strategy parameters. Constructing countermeasure strategy constraints, which include power constraints, frequency constraints, time constraints, and distance constraints. The power constraints are used to limit the power range of the countermeasure strategy, the frequency constraints are used to limit the frequency range of the countermeasure strategy, the time constraints are used to limit the time range of the countermeasure strategy, and the distance constraints are used to limit the distance range of the countermeasure strategy. A global multi-objective optimization function is constructed, which includes a disturbance effect term, an energy consumption term, a disturbance time term, and a disturbance distance term. The terms in the global multi-objective optimization function are weighted and combined using weighting coefficients. The countermeasure strategy interval is discretized into multiple strategy segments, and the global multi-objective optimization function is discretized to construct a global discretized objective function. The global discretized objective function includes the interference effect of the strategy segment, the energy consumption of the strategy segment, the time of the strategy segment, and the distance of the strategy segment. The global discretized objective function is iteratively optimized using a sequential quadratic programming method. By constructing a quadratic approximation of the global multi-objective optimization function and linear constraints, the parameter increments of the control points are obtained by solving the quadratic programming problem. The parameter values ​​of the control points are then updated based on these parameter increments. The optimized countermeasure strategy is smoothed by fifth-order spline interpolation. By maintaining the continuity of the parameter derivative, power derivative, and frequency derivative at the interpolation endpoints, a smooth and continuous countermeasure strategy is generated. Based on the smooth and continuous countermeasure strategy, an interference parameter sequence and corresponding interference control commands are generated.

2. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 1, characterized in that, The process of constructing the generative adversarial network model includes: Acquire real-time signal data, which includes signal frequency, signal strength, signal modulation method, and signal source location; A signal feature space is constructed based on the real-time signal data, and a classification space is constructed based on the distinguishability of the signal features. A multi-objective loss function is constructed based on the signal feature space and the classification space. The multi-objective loss function includes a classification loss term, a feature matching loss term, and an adversarial loss term. A dual discriminator structure is constructed, which includes a main discriminator and an auxiliary discriminator. Both the main discriminator and the auxiliary discriminator include an input layer, three hidden layers, and an output layer. The generative adversarial network model is trained using a batch training method; The signal identification results are output based on the trained generative adversarial network model. The signal identification results include signal category, signal source location, and signal threat level.

3. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 2, characterized in that, The specific components of the multi-objective loss function include: The classification loss term is calculated using the cross-entropy between the signal features and the classification label; The feature matching loss term is calculated using the Euclidean distance between the signal features and the generated features; The adversarial loss term is calculated by the KL divergence between the probability distribution output by the discriminator and the true distribution.

4. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 2, characterized in that, The dual discriminator structure specifically includes: The dimension of the input layer is the same as the dimension of the signal feature space; The hidden layer uses the LeakyReLU activation function; The output layer has the same dimension as the classification space and uses the Softmax activation function. A joint loss function is constructed based on the main discriminator and the auxiliary discriminator, the joint loss function including the main discriminator loss term and the auxiliary discriminator loss term.

5. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 1, characterized in that, The construction process of the signal countermeasure model includes: Construct a multi-objective function for signal countermeasures, the multi-objective function including an interference effect objective function and an energy consumption objective function; The countermeasure strategy constraints are constructed based on the multi-objective function, and the countermeasure strategy constraints include power constraints, frequency constraints, time constraints, and distance constraints. The countermeasure strategy is encoded using a particle swarm optimization algorithm, with each particle containing information on interference power, interference frequency, and interference time. An adaptive inertia weight mechanism and a dynamic learning factor are introduced for iterative optimization; The optimal countermeasure strategy is selected from the Pareto front solution set that satisfies the trade-off between interference effect and energy consumption.

6. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 5, characterized in that, The specific components of the multi-objective function include: The objective function for the interference effect is calculated by multiplying the signal strength attenuation rate by the interference time. The energy consumption objective function is calculated by multiplying the interference power by the interference time; The particle fitness is obtained by weighting the interference effect and energy consumption.

7. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 5, characterized in that, The optimization process of the improved particle swarm optimization algorithm includes: The adaptive inertia weight mechanism employs a time-varying inertia weight factor, which exhibits a linear decay as the number of iterations increases. The dynamic learning factor increases linearly with the number of iterations. Based on the adaptive inertia weight mechanism and the dynamic learning factor, the interference effect and energy consumption of the countermeasure strategy generated in each iteration are evaluated, and the non-dominated solution is added to the Pareto front solution set.

8. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 1, characterized in that, The execution layer performs local signal interference based on the optimal countermeasure strategy, including: A local interference window is constructed based on the current location, current power, and current frequency of the interference device. The size of the local interference window is adaptively adjusted by a power adaptive coefficient, and a positive correlation is established between the size of the local interference window and the magnitude of the current power. A local signal environment model is constructed using multi-source sensor data. The sensor data is transformed to obtain signal data in the local coordinate system. The signal intensity distribution map is updated based on the signal data. The intensity value of each signal point is calculated using a logarithmic probability update method. The Kalman filter algorithm is used to track the dynamic signal source. The state vector of the signal source is predicted by the state prediction equation. The predicted state is updated based on the measurement data to obtain the precise position and motion information of the signal source. An interference optimization model is constructed, and the kinematic equation of the interference device is used as the state equation. The state equation includes position coordinates, power and frequency. State constraints and dynamic constraints are constructed. The state constraints are used to limit the range of values ​​for position coordinates and power. The dynamic constraints are used to limit the range of values ​​for frequency change rate, power change rate and time change rate. A multi-objective cost function is constructed, which includes a reference interference effect term, a signal source avoidance term, an interference smoothing term, and an energy consumption term. The terms in the multi-objective cost function are weighted and combined using weighting coefficients. The gradient projection method is used to optimize the multi-objective cost function, calculate the gradient of the cost function with respect to the control quantity, project the updated control quantity to the feasible region, and determine the optimal step size through backtracking search.

9. The method for intelligent detection and countermeasure of UAV signals based on deep learning according to claim 1, characterized in that, The feedback layer, based on an adaptive filtering algorithm, enables real-time evaluation and adjustment of the countermeasures effect, including: An interference effect evaluation model is established, which includes a signal strength attenuation equation and an interference time equation. The signal strength attenuation equation includes an interference power term, an interference frequency term, and an interference time term. The interference time equation includes an interference power term and an interference frequency term. The interference effect evaluation model is constructed as a state-space expression. The state vector of the state-space expression includes the signal strength attenuation rate, interference time, and interference distance. The control vector of the state-space expression includes the interference power and interference frequency. The state-space expression is linearized, and the Jacobian matrix of the system state equation with respect to the state vector and control vector is calculated to construct a linearized prediction model. A cost function for predicting interference effects is constructed. The cost function includes an interference effect error term, a control quantity penalty term, and a control increment penalty term. The penalty terms are weighted and combined using a weight matrix. State constraints are constructed, including signal strength attenuation rate constraints and interference time constraints. Control constraints are constructed, including interference power constraints and interference frequency constraints. Control increment constraints are constructed, including interference power increment constraints and interference frequency increment constraints. The prediction cost function is transformed into the standard form of a quadratic programming problem, and the quadratic form matrix and the coefficients of the linear terms are calculated to construct the inequality constraint matrix and the equality constraint matrix. The quadratic programming problem is solved using the interior point method. Based on the optimized solution, the control variables are mapped. The total interference power is allocated to each interference device through the power allocation matrix, and the interference frequency is allocated to each interference device through the frequency allocation matrix. The output of the interference device is limited, the interference power is limited according to the power characteristics, the interference frequency is limited according to the frequency characteristics, and the final interference control command is generated.

Images