Unmanned aerial vehicle signal adaptive interference system based on multi-mode perception

Abstract

The invention discloses an unmanned aerial vehicle signal adaptive interference system based on multi-modal sensing, and belongs to the technical field of unmanned aerial vehicle signal adaptive interference systems. The data fusion and processing module is used for carrying out fusion processing on the features output by the multi-mode sensing module to realize identification and positioning of the unmanned aerial vehicle, and the adaptive interference strategy generation module is used for generating an adaptive interference strategy according to the identification and positioning result of the unmanned aerial vehicle. And the interference signal transmitting module is used for transmitting an interference signal according to the generated interference strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to an assembly positioning mechanism, in particular to a multi-modal perception-based unmanned aerial vehicle signal adaptive jamming system, belonging to the technical field of assembly positioning mechanisms. BACKGROUND

[0002] With the rapid development of unmanned aerial vehicle technology, unmanned aerial vehicles are increasingly widely used. However, the misuse of unmanned aerial vehicles has also brought a series of safety problems. Therefore, it is crucial to effectively control and counter unmanned aerial vehicles.

[0003] Currently, common unmanned aerial vehicle countermeasures include radio frequency jamming, GPS jamming, laser strikes, etc. Among them, radio frequency jamming is a commonly used method, which destroys the communication link between the unmanned aerial vehicle and the remote controller or the navigation positioning signal of the unmanned aerial vehicle by emitting jamming signals, thereby achieving control over the unmanned aerial vehicle. However, existing radio frequency jamming technology has some limitations: Existing jamming systems often rely solely on a single signal feature to identify and jam unmanned aerial vehicles. In complex electromagnetic environments, they are easily disturbed by other signals, leading to misjudgment and missed judgment. For example, in urban environments, there are a large number of wireless communication signals, broadcast television signals, etc. These signals may be confused with unmanned aerial vehicle signals, making it difficult for the jamming system to accurately identify unmanned aerial vehicle signals.

[0004] Traditional jamming strategies are usually pre-set and cannot be dynamically adjusted according to the real-time state of the unmanned aerial vehicle and environmental changes. Different types of unmanned aerial vehicles may have different communication protocols, frequency ranges, and anti-jamming capabilities. Fixed jamming strategies are difficult to effectively jam all types of unmanned aerial vehicles. Moreover, when unmanned aerial vehicles take some evasive measures (such as changing flight altitude, speed, communication frequency, etc.), traditional jamming systems cannot respond in time, leading to jamming failure.

[0005] Multiple unmanned aerial vehicles are difficult to coordinate: In the face of multiple unmanned aerial vehicles working together, existing jamming systems are difficult to effectively jam multiple unmanned aerial vehicles at the same time. Multiple unmanned aerial vehicles may use distributed communication, ad hoc networking, etc. to collaborate to complete tasks. Existing jamming systems cannot comprehensively perceive and analyze these coordinated unmanned aerial vehicles, making it difficult to develop effective jamming strategies for multiple unmanned aerial vehicle coordination.

[0006] Therefore, a more intelligent and efficient unmanned aerial vehicle signal adaptive jamming system is needed, which can comprehensively utilize multiple perception information to accurately identify and locate unmanned aerial vehicles, and adaptively adjust the jamming strategy according to the real-time state of the unmanned aerial vehicle and environmental changes, to effectively counter unmanned aerial vehicles. SUMMARY

[0007] The main purpose of the present application is to provide a multi-modal perception-based unmanned aerial vehicle signal adaptive jamming system to solve the single-modal perception limitation, lack of adaptability of jamming strategies, and difficulty in coordinated response to multiple unmanned aerial vehicles in existing unmanned aerial vehicle countermeasure technologies. By fusing multiple perception information, accurate identification and positioning of unmanned aerial vehicles are realized, and jamming strategies can be dynamically adjusted according to the real-time state of unmanned aerial vehicles and environmental changes, thereby improving the jamming effect and countermeasure capability of unmanned aerial vehicles.

[0008] The object of the present application can be achieved by adopting the following technical solutions: The multi-modal perception-based unmanned aerial vehicle signal adaptive jamming system comprises a multi-modal perception module, a data fusion and processing module, an adaptive jamming strategy generation module, and a jamming signal transmission module. The multi-modal perception module is used for collecting multi-modal signals of unmanned aerial vehicles and extracting features. The data fusion and processing module is used for fusing and processing the features output by the multi-modal perception module to realize identification and positioning of unmanned aerial vehicles. The adaptive jamming strategy generation module is used for generating adaptive jamming strategies according to the identification and positioning results of unmanned aerial vehicles. The jamming signal transmission module is used for transmitting jamming signals according to the generated jamming strategies.

[0009] Preferably, the multi-modal perception module comprises a radio frequency signal perception unit, a visual perception unit, and an acoustic perception unit. The radio frequency signal perception unit adopts a wide-band radio frequency receiving antenna array for receiving radio frequency signals transmitted by unmanned aerial vehicles at different communication frequency bands and extracting frequency, amplitude, phase, and modulation mode features of the radio frequency signals. The visual perception unit is composed of a high-definition camera and a thermal imager for acquiring optical image information and thermal radiation information of unmanned aerial vehicles and identifying shape, color, and size features of the unmanned aerial vehicles. The acoustic perception unit comprises a microphone array composed of multiple microphones for collecting sound signals generated during the flight of unmanned aerial vehicles and extracting frequency spectrum features of the sound signals.

[0010] Preferably, the data fusion and processing module comprises a time synchronization and space registration unit, a multi-modal feature fusion unit, and an unmanned aerial vehicle identification and positioning unit. The time synchronization and space registration unit is used for performing time synchronization processing on the data of the radio frequency signal perception unit, the visual perception unit, and the acoustic perception unit, establishing a unified space coordinate system, and registering the space position information of different modal data. The multi-modal feature fusion unit adopts a multi-modal fusion network based on an attention mechanism to fuse the features of different modal data. The unmanned aerial vehicle recognition and positioning unit inputs the fused multi-modal features into the trained unmanned aerial vehicle recognition model to realize recognition of the unmanned aerial vehicle type, model and identity, and uses a multi-source information fusion positioning algorithm to realize three-dimensional positioning of the unmanned aerial vehicle.

[0011] Preferably, the multi-source information fusion positioning algorithm is a joint positioning algorithm based on trilateration and signal angle of arrival, which realizes positioning by combining the angle of arrival of the radio frequency signal, the position information in the visual image and the propagation time difference of the acoustic signal.

[0012] Preferably, the adaptive interference strategy generation module comprises a target state evaluation model unit, a reinforcement learning algorithm unit and an interference strategy decision unit. The target state evaluation model unit is configured to quantitatively evaluate the threat degree of the unmanned aerial vehicle by using multi-dimensional evaluation indexes according to the position coordinates, motion state and behavior mode information of the unmanned aerial vehicle. The reinforcement learning algorithm unit uses a reinforcement learning algorithm, takes the unmanned aerial vehicle state information output by the target state evaluation model unit as the state input, takes the interference strategy type and interference parameter configuration as the action space, and optimizes the interference strategy by interacting with the environment. The interference strategy decision unit is configured to make a final decision on the interference strategy in combination with the system resource state and task demand according to the interference strategy generated by the reinforcement learning algorithm.

[0013] Preferably, the multi-dimensional evaluation indexes comprise a spatial activity range index, a motion characteristic index and a behavior characteristic index. The spatial activity range index is the ratio of the distance between the unmanned aerial vehicle and the important target to the preset safety distance. The motion characteristic index comprises the speed, acceleration and flight attitude change rate of the unmanned aerial vehicle. The behavior characteristic index is quantified according to the flight trajectory mode of the unmanned aerial vehicle.

[0014] Preferably, the reinforcement learning algorithm is a deep Q network or a policy gradient algorithm; the interference strategy type comprises radio frequency interference, GPS interference and hybrid interference; and the interference parameter configuration comprises the interference signal frequency, power and modulation mode.

[0015] Preferably, the interference signal emission module comprises an interference signal generation unit, a power adjustment and beamforming unit and an interference signal emission unit. The interference signal generation unit is configured to generate corresponding interference signals by a waveform generator according to the interference strategy determined by the adaptive interference strategy generation module. The power adjustment and beamforming unit is configured to use a power amplifier to amplify the power of the interference signals and use an adaptive beamforming technology to adjust the emission beam direction of the interference signals. The interference signal emitting unit is configured to emit interference signals through the emitting antenna.

[0016] Preferably, when the radio frequency interference strategy is adopted, the interference signal generating unit generates interference signals with the same or similar frequency as the unmanned aerial vehicle communication frequency, and adopts noise modulation or sweep frequency modulation mode; when the GPS interference strategy is adopted, a pseudo-satellite signal with the same format as the GPS satellite signal but with different encoding is generated.

[0017] The beneficial technical effects of the present application are: The unmanned aerial vehicle signal adaptive interference system based on multi-modal perception provided by the present application improves the accuracy of unmanned aerial vehicle identification and positioning: by fusing radio frequency signals, vision, and acoustic and other multi-modal perception information, the complementarity of each modal data is fully utilized, which can effectively improve the identification accuracy and positioning accuracy of unmanned aerial vehicles. In complex electromagnetic environments and low visibility and other adverse conditions, the unmanned aerial vehicle can still be accurately identified and positioned, reducing the probability of misjudgment and missed judgment.

[0018] The reinforcement learning algorithm is adopted to automatically generate the optimal interference strategy according to the real-time state of the unmanned aerial vehicle and the environmental changes. It can dynamically adjust the interference strategy for unmanned aerial vehicles of different types and states and complex electromagnetic environments, improve the interference effect, and effectively cope with various evasion measures of unmanned aerial vehicles.

[0019] The multi-modal perception module can simultaneously monitor and perceive multiple unmanned aerial vehicles in the airspace, the data fusion and processing module can comprehensively analyze and process the information of multiple unmanned aerial vehicles, and the adaptive interference strategy generation module can formulate corresponding interference strategies according to the cooperative operation mode and threat level of multiple unmanned aerial vehicles, effectively counteracting the cooperative operation of multiple unmanned aerial vehicles.

[0020] Accurate unmanned aerial vehicle identification and positioning and adaptive interference strategies can avoid false interference on non-unmanned aerial vehicle signals, reduce the false alarm rate of the system, and dynamically adjust the interference strategy and transmitting power according to actual needs, avoiding unnecessary resource waste and improving the working efficiency and resource utilization rate of the system. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 A system diagram of a preferred embodiment of the unmanned aerial vehicle signal adaptive interference system based on multi-modal perception according to the present application. DETAILED DESCRIPTION

[0022] In order to make those skilled in the art more clear and explicit about the technical solutions of the present application, the present application will be further described in detail below in conjunction with the embodiments and drawings, but the implementation manner of the present application is not limited thereto.

[0023] To achieve the above object, the application provides a multi-modal perception-based unmanned aerial vehicle signal adaptive jamming system, which comprises the following main parts: a multi-modal perception module: A radio frequency signal perception unit: a wide-band radio frequency receiving antenna array is adopted, which can receive radio frequency signals emitted by unmanned aerial vehicles at different communication frequency bands (such as 900M, 1.2G, 2.4GHz, 5.8GHz, etc.). By analyzing the frequency, amplitude, phase, modulation mode and other characteristics of the radio frequency signal, the communication link information and part of the identity recognition information of the unmanned aerial vehicle are obtained. For example, unmanned aerial vehicles produced by different manufacturers may use different modulation modes (such as FSK, QAM, etc.), and the type of unmanned aerial vehicle can be preliminarily judged by identifying the modulation mode.

[0024] A visual perception unit: a visual perception system composed of a high-definition camera and a thermal imager is used to monitor the airspace in real time. The high-definition camera is used to obtain optical image information of the unmanned aerial vehicle, and the shape, color, size and other characteristics of the unmanned aerial vehicle are identified by a target detection algorithm (such as YOLO, FasterR-CNN and other algorithms based on deep learning), so as to further determine the model and category of the unmanned aerial vehicle. The thermal imager can realize effective monitoring and tracking of the unmanned aerial vehicle at night or in low-visibility environment by detecting the heat radiation generated by the heat generating components such as the engine of the unmanned aerial vehicle.

[0025] An acoustic perception unit: a microphone array is arranged, which is used to collect sound signals generated during the flight of the unmanned aerial vehicle. Different types of unmanned aerial vehicles have different sound characteristics due to differences in the number of motors, sizes and rotation speeds of propellers, etc. Through spectrum analysis and feature extraction of the sound signal, the type and approximate position of the unmanned aerial vehicle can be identified, providing auxiliary information for subsequent jamming.

[0026] A data fusion and processing module: Time synchronization and space registration: the data from the radio frequency signal perception unit, the visual perception unit and the acoustic perception unit are subjected to time synchronization processing to ensure the consistency of different modal data in time. At the same time, by establishing a unified spatial coordinate system, the spatial position information of different modal data is registered, so that the modal data can be fused and analyzed in the same spatial framework. For example, by using GPS positioning information and camera calibration parameters, the image coordinates of the unmanned aerial vehicle obtained by the visual perception unit are converted into the same spatial coordinates as the radio frequency signal perception unit and the acoustic perception unit.

[0027] Multi-modal feature fusion: Construct a multi-modal feature fusion network using deep learning algorithms (such as multi-modal fusion networks based on attention mechanisms) to fuse the features of different modal data. This network can automatically learn the correlation and complementarity between different modal features, providing more comprehensive and accurate feature representations for subsequent drone identification and positioning. For example, the frequency features of radio frequency signals, the shape features of visual images, and the spectral features of sound signals are fused to form a comprehensive feature vector to describe the state of the drone.

[0028] Drone identification and positioning: Input the fused multi-modal features into the trained drone identification model (such as a classification model based on convolutional neural networks) to accurately identify the type, model, and identity of the drone. At the same time, use multi-source information fusion positioning algorithms (such as joint positioning algorithms based on trilateration and signal angle of arrival) combined with the angle of arrival of radio frequency signals, location information in visual images, and time difference of acoustic signal propagation, etc. to accurately locate the drone in three dimensions.

[0029] Adaptive interference strategy generation module: Target state evaluation model: Based on the drone position coordinates, motion state (speed, acceleration, flight attitude, etc.), and behavior pattern (circling, straight flight, hovering, etc.) information output by the drone identification and positioning module, construct a target state evaluation model. This model uses multi-dimensional evaluation indicators such as spatial activity range indicators, motion characteristic indicators, and behavior feature indicators to quantitatively evaluate the threat level of the drone. For example, when the drone approaches important facilities or densely populated areas, its spatial activity range indicator will decrease, and the threat level will correspondingly increase.

[0030] Reinforcement learning algorithm: Use reinforcement learning algorithms (such as deep Q networks, policy gradient algorithms, etc.) to design intelligent interference strategies. The drone state information output by the target state evaluation model is used as the state input of the reinforcement learning environment, and the interference strategy type (such as radio frequency interference, GPS interference, mixed interference, etc.) and interference parameter configuration (such as interference signal frequency, power, modulation mode, etc.) are used as the action space. Through continuous interaction with the environment, the interference strategy is optimized according to the feedback reward signal (such as whether the interference is successful, the distance change between the drone and the target area, etc.), so that the interference system can automatically generate the optimal interference strategy according to the real-time state of the drone and environmental changes.

[0031] Interference strategy decision: According to the interference strategy generated by the reinforcement learning algorithm, combined with the current system resource state (such as interference signal transmission power limit, available frequency band, etc.) and task demand (such as focus protection on specific area, priority interference on different types of unmanned aerial vehicles, etc.), the final decision of interference strategy is made. For example, if the current system has limited transmission power, but the threat level of unmanned aerial vehicles is high, then the interference strategy with higher power efficiency is preferred; if the task requires focus protection on a specific area, then the unmanned aerial vehicles entering this area are given priority for interference measures.

[0032] Interference signal transmission module: Interference signal generation: According to the interference strategy determined by the adaptive interference strategy generation module, the corresponding interference signal is generated by the waveform generator. If the radio frequency interference strategy is adopted, the interference signal with the same or similar frequency as the unmanned aerial vehicle communication frequency is generated, and the unmanned aerial vehicle communication signal is interfered by modulation methods (such as noise modulation, sweep modulation, etc.); if the GPS interference strategy is adopted, the pseudo-satellite signal that can interfere with the unmanned aerial vehicle GPS positioning signal is generated.

[0033] Power adjustment and beamforming: The power amplifier is used to amplify the interference signal to meet the interference demand. At the same time, the adaptive beamforming technology is adopted to dynamically adjust the transmission beam direction of the interference signal according to the real-time position and direction of the unmanned aerial vehicle, so that the interference signal can be concentrated on the target unmanned aerial vehicle, improving the power utilization rate of the interference signal and the interference effect. For example, by controlling the phase and amplitude of each element in the antenna array, the beamforming of the interference signal is realized, so that the interference beam can accurately track the movement of the unmanned aerial vehicle.

[0034] Interference signal transmission: The generated interference signal is transmitted through the interference signal transmission antenna to interfere with the target unmanned aerial vehicle. The transmission antenna can choose different types according to the actual application scene, such as directional antenna, omnidirectional antenna, etc. When interference is needed on unmanned aerial vehicles in a specific direction, directional antenna can improve the pertinence of interference; when interference is needed on unmanned aerial vehicles in a large area, omnidirectional antenna is more appropriate.

[0035] Implementation of multi-modal perception module: Radio frequency signal perception unit: A radio frequency antenna array with wide frequency band receiving capability is selected, such as an array composed of multiple log-periodic antennas, which can cover common unmanned aerial vehicle communication frequency bands. The received radio frequency signals are amplified, filtered and down-converted by the radio frequency front-end circuit, and then converted into digital signals and sent to the digital signal processing unit. The digital signal processing unit uses fast Fourier transform (FFT) and other algorithms to analyze the frequency spectrum of the signal, extracts the frequency, amplitude, phase and other characteristics of the signal, and uses modulation recognition algorithms to identify the modulation mode of the signal.

[0036] Visual Perception Unit: High-definition cameras and thermal imagers are installed on a rotatable gimbal to monitor different directions of airspace. High-definition cameras are selected with high resolution and large field of view. Image data collected by the cameras is transmitted to the computer through an image acquisition card. A deep learning framework such as TensorFlow or PyTorch is used to train a target detection model to identify and track drones in images. Thermal imagers transmit thermal imaging data to the computer through a serial port or network interface. Temperature threshold-based and morphological processing methods are used to detect and track thermal targets of drones.

[0037] Acoustic Perception Unit: Multiple microphones are arranged in a geometric layout to form a microphone array, such as a circular array or a linear array. The sound signals collected by the microphone array are converted into digital signals by an audio acquisition card and sent to the computer. Beamforming algorithms such as delay-and-sum beamforming are used to process sound signals to estimate the direction of the sound source of the drone. Meanwhile, sound feature extraction algorithms such as Mel Frequency Cepstral Coefficients (MFCC) are used to extract features from sound signals for drone type identification. Data Fusion and Processing Module Implementation: Time Synchronization and Spatial Registration: High-precision clock sources such as GPS-synchronized clocks are used to synchronize the clocks of the radio frequency signal perception unit, visual perception unit, and acoustic perception unit, ensuring the consistency of data collected by each unit in time. In terms of spatial registration, the camera in the visual perception unit is calibrated to obtain the camera's intrinsic and extrinsic matrices. The angle of arrival information obtained by the radio frequency signal perception unit and the time difference of arrival information obtained by the acoustic perception unit are combined with the camera's extrinsic matrix to convert the spatial position information of different modal data to a unified coordinate system.

[0038] Multi-modal Feature Fusion: A multi-modal feature fusion network based on attention mechanism is constructed. The network includes three branches corresponding to the extraction of radio frequency signal features, visual features, and acoustic features. Each branch uses a convolutional neural network (CNN) to extract features from its own modal data, resulting in different levels of feature maps. Then, the attention mechanism module calculates the weights between different modal features, and the weighted features are fused. Finally, the fused features are input into the fully connected layer for classification and regression, achieving drone identification and positioning.

[0039] Drone identification and localization: A large amount of drone sample data (including different types, models of drones in different flight states, and multi-modal data) is used to train the drone identification model. During the training process, the cross-entropy loss function and the stochastic gradient descent algorithm are used to optimize the parameters of the model. In terms of localization, a joint positioning algorithm based on trilateration and signal angle of arrival is used. According to the signal angle of arrival information obtained by the radio frequency signal perception unit and the position information of multiple receiving nodes, the position coordinates of the drone are determined through geometric calculation. At the same time, combined with the position information obtained by the visual perception unit and the acoustic perception unit, the positioning results are optimized and corrected.

[0040] Implementation of adaptive interference strategy generation module: Target state evaluation model: The spatial activity range index is defined as the ratio of the distance between the drone and the important target to the preset safety distance; the motion characteristic index includes the speed, acceleration, and flight attitude change rate of the drone; the behavior characteristic index is quantified according to the flight trajectory mode of the drone (such as straight flight, hovering, round-trip flight, etc.). These evaluation indexes are used as input to build a multi-layer perception (MLP) model as the target state evaluation model. A large amount of drone flight data is used to train the model, so that it can accurately evaluate the threat level of the drone.

[0041] Reinforcement learning algorithm: A deep Q network (DQN) is used as the reinforcement learning algorithm. The drone state information output by the target state evaluation model is used as the state input of DQN, and the interference strategy type and interference parameter configuration are used as the action space. In the training process, a reasonable reward function is set, such as giving a positive reward when the interference successfully makes the drone return or land, and giving a negative reward when the interference fails or the drone approaches the important target. Through continuous interaction with the environment, the network parameters of DQN are updated according to the reward signal, so that the interference strategy is gradually optimized.

[0042] Interference strategy decision: A interference strategy library is established to store various interference strategies that have been trained and verified. In actual application, according to the interference strategy index generated by the reinforcement learning algorithm, the corresponding interference strategy is selected from the interference strategy library. At the same time, considering the current system resource state (such as transmission power, available frequency band, etc.) and task demand, the interference strategy is appropriately adjusted and optimized. For example, if the available frequency band is limited, the strategy with better interference effect in the frequency band is preferred; if the task requires special protection for a certain area, a more powerful interference strategy is used for the drone entering the area.

[0043] Implementation of interference signal transmission module: Interference signal generation: If the radio frequency interference strategy is adopted, the corresponding interference signal is generated by the waveform generator according to the selected interference strategy. For example, when noise modulation interference is adopted, the waveform generator generates a Gaussian white noise signal and modulates it to the same carrier as the unmanned aerial vehicle communication frequency. If the GPS interference strategy is adopted, the waveform generator generates a pseudo-satellite signal with the same format as the GPS satellite signal but with different coding to interfere with the GPS positioning system of the unmanned aerial vehicle.

[0044] Power adjustment and beamforming: The interference signal is power amplified by a power amplifier, and the amplification factor of the power amplifier is adjusted according to the interference strategy and the distance of the target unmanned aerial vehicle. In terms of beamforming, the beamforming technology based on adaptive antenna array is adopted. By controlling the phase and amplitude of each element in the antenna array, the beam of the interference signal is directed to the target unmanned aerial vehicle. For example, the least mean square error (LMS) algorithm is used to adjust the weight of the element according to the real-time position information of the unmanned aerial vehicle, realizing the adaptive tracking of the interference beam.

[0045] Interference signal transmission: According to the actual application scene, select the appropriate transmitting antenna. When interfering with unmanned aerial vehicles in a small area, directional antennas such as Yagi antennas can be selected to concentrate the interference signal to the target direction. When protecting a large area, omnidirectional antennas such as dipole antennas are used to make the interference signal cover the entire area. The transmitting antenna is connected to the power amplifier through the radio frequency cable, and the amplified interference signal is transmitted out to interfere with the target unmanned aerial vehicle.

[0046] The above is only a further embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art within the scope disclosed by the present application, according to the technical solution and concept of the present application, makes equivalent replacement or change, belongs to the protection scope of the present application.

Claims

1. The UAV signal adaptive jamming system based on multimodal perception is characterized by: It includes a multimodal perception module, a data fusion and processing module, an adaptive interference strategy generation module, and an interference signal transmission module; The multimodal sensing module is used to collect multimodal signals of the UAV and extract features; The data fusion and processing module is used to fuse the features output by the multimodal perception module to achieve the recognition and positioning of the drone; The adaptive interference strategy generation module is used to generate an adaptive interference strategy based on the identification and positioning results of the UAV; The interference signal transmitting module is used to transmit an interference signal according to the generated interference strategy.

2. The multimodal sensing-based UAV signal adaptive jamming system according to claim 1 is characterized by: The multimodal perception module includes a radio frequency signal perception unit, a visual perception unit and an acoustic perception unit; The radio frequency signal sensing unit adopts a wide-band radio frequency receiving antenna array to receive radio frequency signals transmitted by drones in different communication frequency bands and extract the frequency, amplitude, phase and modulation characteristics of the radio frequency signals; The visual perception unit is composed of a high-definition camera and a thermal imager, which is used to obtain optical image information and thermal radiation information of the drone and identify the shape, color, and size characteristics of the drone; The acoustic sensing unit is composed of a microphone array composed of multiple microphones, which is used to collect the sound signals generated during the flight of the drone and extract the spectral characteristics of the sound signals.

3. The multimodal sensing-based UAV signal adaptive jamming system according to claim 2 is characterized by: The data fusion and processing module includes a time synchronization and spatial registration unit, a multimodal feature fusion unit, and a UAV identification and positioning unit; The time synchronization and spatial registration unit is used to perform time synchronization processing on the data of the radio frequency signal perception unit, the visual perception unit and the acoustic perception unit, and to establish a unified spatial coordinate system to align the spatial position information of the different modal data; The multimodal feature fusion unit adopts a multimodal fusion network based on an attention mechanism to fuse features of data from different modalities; The drone identification and positioning unit inputs the fused multimodal features into a trained drone identification model to identify the type, model and identity of the drone, and uses a positioning algorithm based on multi-source information fusion to perform three-dimensional positioning of the drone.

4. The multimodal sensing-based UAV signal adaptive jamming system according to claim 3 is characterized by: The multi-source information fusion positioning algorithm is a joint positioning algorithm based on trilateration and signal arrival angle, which combines the arrival angle of the radio frequency signal, the position information in the visual image and the propagation time difference of the acoustic signal to achieve positioning.

5. The multimodal sensing-based UAV signal adaptive jamming system according to claim 4 is characterized by: The adaptive interference strategy generation module includes a target state evaluation model unit, a reinforcement learning algorithm unit and an interference strategy decision unit; The target state assessment model unit is used to quantitatively assess the threat level of the drone using multi-dimensional assessment indicators based on the drone's position coordinates, motion state, and behavior pattern information; The reinforcement learning algorithm unit adopts a reinforcement learning algorithm, takes the drone state information output by the target state evaluation model unit as state input, and the interference strategy type and interference parameter configuration as action space, and optimizes the interference strategy by interacting with the environment; The interference strategy decision unit is used to make a final decision on the interference strategy based on the interference strategy generated by the reinforcement learning algorithm and in combination with the system resource status and task requirements.

6. The multimodal sensing-based UAV signal adaptive jamming system according to claim 5 is characterized by: The multi-dimensional evaluation indicators include spatial activity range indicators, movement characteristics indicators and behavioral characteristics indicators; The spatial activity range indicator is the ratio of the distance between the drone and the important target to the preset safety distance; The motion characteristic indicators include the speed, acceleration and flight attitude change rate of the UAV; The behavioral characteristic index is quantified according to the flight trajectory pattern of the UAV.

7. The multimodal sensing-based UAV signal adaptive jamming system according to claim 6 is characterized by: The reinforcement learning algorithm is a deep Q network or a policy gradient algorithm; the interference strategy types include radio frequency interference, GPS interference, and mixed interference; the interference parameter configuration includes interference signal frequency, power, and modulation mode.

8. The multimodal sensing-based UAV signal adaptive jamming system according to claim 7 is characterized by: The interference signal transmission module includes an interference signal generation unit, a power adjustment and beamforming unit and an interference signal transmission unit; The interference signal generating unit is configured to generate a corresponding interference signal by a waveform generator according to the interference strategy determined by the adaptive interference strategy generating module; The power adjustment and beamforming unit is used to amplify the power of the interference signal using a power amplifier and adjust the transmission beam direction of the interference signal using an adaptive beamforming technology; The interference signal transmitting unit is used to transmit the interference signal through the transmitting antenna.

9. The multimodal sensing-based UAV signal adaptive jamming system according to claim 8 is characterized by: When the radio frequency interference strategy is adopted, the interference signal generation unit generates an interference signal with the same or similar frequency as the drone communication frequency, using noise modulation or sweep frequency modulation; when the GPS interference strategy is adopted, a pseudo-satellite signal with the same format as the GPS satellite signal but with different coding is generated.

Images