An unmanned aerial passive signal reconnaissance positioning adaptive regulation method and system

Abstract

The application discloses a kind of unmanned aerial vehicle borne passive signal reconnaissance positioning adaptive regulation method and system, belong to signal processing technology, environmental perception technology and electromagnetic spectrum monitoring technical field;Method includes: real-time acquisition communication / radar, clutter and interference signal, fusion unmanned aerial vehicle IMU / GNSS data, complete electromagnetic environment cognition by CNN-LSTM;Multi-objective optimization function is constructed, to convex optimization is combined with reinforcement learning output optimal receiving parameter;Adaptive configuration radio frequency front end, through STAP and cognitive filtering suppresses clutter interference;Based on deep learning, signal separation, identification, direction finding and positioning are realized;Quantitative evaluation and closed loop iteration optimization;The application realizes the adaptive acquisition of weak communication / radar signal under complex electromagnetic environment, accurate identification, high-precision direction finding and rapid positioning, improves monitoring robustness and adaptability, adapts unmanned aerial vehicle dynamic flight and strong interference scene, and can be widely applied in low radio monitoring and other fields.

Description

Technical Field

[0001] This invention belongs to the fields of signal processing technology, environmental perception technology and electromagnetic spectrum monitoring technology, and particularly relates to an adaptive control method and system for unmanned aerial vehicle (UAV) passive signal reception and positioning. Background Technology

[0002] With the increasing complexity of the low-altitude economy and electromagnetic warfare environment, drones equipped with electromagnetic spectrum monitoring equipment have become an important means of long-range monitoring, acquisition, identification, direction finding, and positioning of communication and radar signals due to their advantages such as mobility, rapid deployment, and no personnel risk. Current radio monitoring faces challenges such as dense spectrum, complex signal types, strong interference, and high dynamism: it includes various communication signals (FM, LTE, 5G, frequency hopping communication) as well as radar transmission signals from various systems (LFM, phase coding, pulse Doppler); simultaneously, there are ground clutter, electromagnetic interference, co-channel and adjacent-channel interference, and channel distortion caused by the attitude motion of the drone platform. This makes it difficult for traditional fixed-parameter receivers to achieve weak signal acquisition, signal type identification, high-precision direction finding, and rapid positioning, resulting in problems such as missed detections, misidentification, large direction finding errors, and low positioning accuracy.

[0003] Existing UAV-borne monitoring equipment mostly uses fixed frequency bands, fixed filtering, and fixed direction-finding algorithms, lacking real-time spectrum perception, dynamic adaptive reception parameters, deep signal feature recognition, online optimization of direction-finding / positioning algorithms, and end-to-end closed-loop iteration capabilities. Moreover, most are designed only for single communication or single radar signals and cannot be compatible with mixed communication / radar signal scenarios. Furthermore, the lack of deep integration of UAV platform attitude, position, and velocity information with electromagnetic signal data results in poor robustness and weak adaptability in complex environments, making it difficult to meet the actual needs of signal monitoring, reconnaissance, and positioning in complex electromagnetic environments.

[0004] Therefore, there is an urgent need to develop an unmanned aerial vehicle (UAV) electromagnetic spectrum sensing and communication / radar signal monitoring direction finding and positioning method and system to achieve real-time full-spectrum sensing, adaptive signal acquisition, accurate identification of communication / radar signals, high-precision direction finding and positioning, and closed-loop adaptive optimization, adapting to the dynamic flight of UAVs and complex electromagnetic interference scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive control method and system for unmanned aerial vehicle (UAV) passive signal reception and positioning, addressing the problems of weak adaptive capability, poor clutter and interference suppression, low signal recognition accuracy, large direction finding and positioning errors, and lack of closed-loop optimization mechanisms in existing technologies. This invention constructs a full-link cognitive closed loop to achieve coordinated operation of electromagnetic environment perception, receiver parameter decision-making, adaptive receiver processing, signal recognition, and direction finding and positioning, thereby improving the accuracy and robustness of communication / radar signal monitoring, reconnaissance, and positioning in complex electromagnetic environments.

[0006] To achieve the above objectives, the present invention provides an adaptive control method for unmanned aerial vehicle (UAV) passive signal reception and positioning, comprising the following steps: S1. The UAV-borne cognitive radar collects communication signals, radar signals, clutter signals, and electromagnetic interference signals in the detection area in real time, and simultaneously integrates attitude, position, and velocity information output by the UAV's inertial measurement unit (IMU) and global navigation satellite system (GNSS) module to construct a multi-dimensional perception dataset; based on the perception dataset, electromagnetic environment and signal features are extracted, and clutter type, interference intensity, interference level, signal type, and modulation method are identified to complete the preliminary understanding of the electromagnetic environment situation and potential target signals; S2. A multi-objective optimization function is constructed with the optimization objectives of maximizing the weak signal detection probability, maximizing the interference suppression ratio, maximizing the direction finding accuracy, and minimizing the false alarm rate. Combining the electromagnetic environment situation, preliminary signal characteristics, and the receiver's historical operating status, the optimal combination of receiving parameters is dynamically output through the optimization algorithm. S3. Based on the optimal receiving parameters, drive the receiver front-end to complete frequency band adaptive switching, real-time adjustment of filter parameters, adaptive control of receiving gain, and synchronously adjust multi-channel sampling and array preprocessing parameters; perform space-time joint preprocessing on the received signal to suppress clutter and electromagnetic interference. S4. Perform precise clutter interference modeling and signal separation on the preprocessed signal from S3, and extract communication / radar signal features; dynamically adjust the identification and direction finding thresholds based on the current clutter and interference intensity to complete communication / radar signal identification, high-precision direction finding and radiation source localization under complex backgrounds, and output target information. S5. Quantitatively evaluate the signal recognition effect, clutter interference suppression effect, direction finding accuracy, positioning error and false alarm rate; feed the evaluation results back to step S2, iteratively optimize the receiving parameters, and form a cognitive closed loop of "spectrum sensing - parameter decision - adaptive reception - identification, direction finding and positioning - feedback optimization".

[0007] Preferably, the multi-dimensional sensing dataset in S1 includes the time-domain characteristics, frequency-domain characteristics, polarization characteristics, and modulation characteristics of communication / radar signals; the amplitude distribution and power spectrum characteristics of clutter; the frequency, power, and modulation characteristics of electromagnetic interference; and the attitude angle, flight speed, and altitude information of the UAV. Clutter types include ground clutter, sea clutter, and meteorological clutter; interference types include co-channel interference, adjacent-channel interference, pulse interference, and suppression interference; and signal types include communication signals and radar signals.

[0008] Preferably, S1 uses a deep neural network for feature extraction and classification, and adopts an adversarial learning initialization strategy to improve the recognition accuracy of signals, clutter and interference under low signal-to-noise ratio conditions. The deep neural network adopts a fusion architecture of convolutional neural network (CNN) and long short-term memory network (LSTM). CNN is used to extract local features in the time and frequency domains of the signal, LSTM is used to capture the temporal correlation of the signal, and the adversarial learning initialization strategy generates simulated signal, clutter and interference samples under low signal-to-noise ratio conditions through generative adversarial network (GAN) to pre-train the neural network.

[0009] Preferably, the specific content of S2 is as follows: S21 Construct a multi-objective optimization function The expression is as follows: ; In the formula, For weak signal detection performance indicators; For interference suppression performance indicators; For direction finding accuracy indicators; The false alarm rate is an indicator. To receive the parameter vector; satisfying ; The constraints of the multi-objective optimization function include bandwidth limitations, gain limitations, dynamic range limitations, power consumption limitations, and flight attitude constraints of the UAV-borne receiver. The receiver parameter combination includes operating frequency band, intermediate frequency bandwidth, filter coefficient, receiver gain, sampling rate, and direction finding baseline configuration; it supports multi-band adaptive switching and dynamic adjustment of multiple filter modes; S22 employs a decision-making mechanism combining convex optimization and reinforcement learning, transforming a non-convex optimization problem into a fast-solvable convex optimization problem. It also achieves online adaptive adjustment of waveform parameters through an environment state-action-reward mapping, as shown in the following expression: ; In the formula, The transformed convex optimization objective function; This is the negative form of the original objective function; The regularization coefficient is used. It is a regular term; The method employs an iterative solution combining the MM algorithm and the Dinkelbach algorithm, and constructs an environment state-action-reward mapping through reinforcement learning; the receiving parameters are then adaptively adjusted online through this environment state-action-reward mapping.

[0010] Preferably, in S3, frequency band switching, filtering adjustment, and gain control are performed through an adaptive RF front-end; the receiver uses space-time adaptive processing STAP combined with a cognitive filtering algorithm to perform preliminary suppression of clutter and interference; the cognitive filtering algorithm, combined with a channel estimation method based on cosine similarity constraints, reduces filtering errors under low signal-to-noise ratio; the STAP technology estimates the covariance matrix through training basis to shape an adaptive receiving beam.

[0011] Preferably, in S4, a stochastic transfer function is used to establish a clutter interference model, and a deep learning algorithm with an attention mechanism is used to separate communication signals, radar signals, clutter, and interference. The stochastic transfer function combines the Green's function impulse response principle to perform accurate clutter modeling, and its expression is as follows: ; In the formula, This is a clutter signal; It is the Green's function impulse response; For random excitation sources; To observe noise; Signal recognition, direction finding, and localization are achieved by employing deep learning signal recognition, array interferometer / spatial spectrum estimation direction finding, and multi-station time difference / frequency difference positioning technology. The detection and recognition threshold is adjusted in real time according to the complexity of the environment.

[0012] Preferably, the quantitative evaluation indicators in S5 include clutter suppression ratio, interference suppression ratio, signal recognition accuracy, direction finding accuracy, positioning error, and false alarm rate; preset performance thresholds for clutter suppression ratio, interference suppression ratio, signal recognition accuracy, direction finding accuracy, positioning error, and false alarm rate; when the evaluation indicators do not reach the preset thresholds, a new round of receiver parameter optimization is immediately started, and the closed-loop iteration cycle is dynamically adjusted according to the complexity of the environment.

[0013] An adaptive control system for passive signal reception and positioning on an unmanned aerial vehicle (UAV), employing any one of the aforementioned adaptive control methods for passive signal reception and positioning on an UAV, comprises: an electromagnetic sensing module, an adaptive decision-making module, a receiving preprocessing module, an identification, direction finding, and positioning module, a closed-loop feedback module, and a data storage module; each module achieves real-time data interaction through a high-speed data bus. The electromagnetic sensing module is used to receive communication / radar signals, clutter, interference signals, and UAV status information, and to extract electromagnetic environment and signal features. Specifically, it includes a multi-channel receiving antenna unit, a radio frequency acquisition unit, a data fusion unit, and a feature extraction and recognition unit. The multi-channel receiving antenna unit is used to receive electromagnetic signals, the radio frequency acquisition unit is used for high-speed sampling, the data fusion unit is used to fuse the status information output by the UAV's IMU and GNSS modules to construct a multi-dimensional sensing dataset, and the feature extraction and recognition unit uses a deep neural network to complete feature extraction and signal, clutter, and interference recognition.

[0014] The adaptive decision-making module is used to construct a multi-objective optimization function and dynamically decide the optimal receiving parameters. It includes a feature recognition unit, a multi-objective optimization unit, and a reinforcement learning unit. The feature recognition unit uses a deep neural network model to extract and recognize features from a multi-dimensional sensing dataset and outputs preliminary characteristics of the electromagnetic environment and signals. The multi-objective optimization unit constructs a multi-objective optimization function and solves for the optimal combination of receiving parameters through a convex optimization algorithm. The reinforcement learning unit combines historical data and real-time feedback information to optimize the receiving parameter decision-making strategy.

[0015] The receiving preprocessing module is used to perform adaptive reception, beamforming, and signal preprocessing based on the receiving parameters. It includes an adaptive radio frequency unit and a cognitive preprocessing unit. The adaptive radio frequency unit includes a reconfigurable filter amplifier and a high-speed sampler, which are used to achieve adaptive frequency band, filtering, gain configuration, and signal acquisition based on the optimal receiving parameters output by the optimization decision module. The cognitive preprocessing unit includes a signal preprocessing module, which is used to acquire electromagnetic signals, perform preliminary filtering and preprocessing, and then transmit them to the identification, direction finding, and positioning module.

[0016] The identification, direction finding, and positioning module is used for signal separation, communication / radar signal identification, high-precision direction finding, and radiation source positioning. It employs technologies such as accurate clutter interference modeling, deep learning separation and identification, array direction finding, and track fusion positioning to refine the pre-processed signal, extract signal features, and complete accurate identification of communication / radar signals, estimation of incoming wave direction, and radiation source positioning. It outputs signal type, frequency, power, azimuth angle, and latitude and longitude location information.

[0017] The closed-loop feedback module is used for performance evaluation and feedback iteration to form a cognitive closed loop; it collects the identification results, direction finding and positioning accuracy, and clutter and interference suppression effects of the identification and direction finding and positioning module, constructs a performance evaluation index system, evaluates the adaptability of the current receiving parameters, and transmits the evaluation feedback information to the optimization decision module to start the parameter iteration optimization process.

[0018] The data storage module is used to store sensing data, historical parameters, recognition results, direction finding and positioning data, and system logs; it adopts a high-speed storage chip to support real-time writing and reading of data, providing data support for optimization decisions and iterative upgrades.

[0019] Preferably, the receiving preprocessing module adopts a multi-element interferometer / phased array receiving antenna, which supports adaptive receiving beamforming and realizes dynamic adjustment of receiving direction and receiving range; the reconfigurable filter amplifier supports real-time switching of multiple frequency bands, multiple filtering modes, and multiple gain levels; The adaptive decision-making module adopts a heterogeneous computing architecture of FPGA+DSP. The FPGA is used to realize high-speed signal processing and fast calculation of received parameters, while the DSP is used to implement complex optimization algorithms and reinforcement learning strategies. The heterogeneous computing architecture realizes parallel computing of feature recognition, multi-objective optimization and reinforcement learning, ensuring the real-time performance of received parameter optimization decisions.

[0020] Preferably, the system also includes a human-machine collaborative decision-making module and a spectrum situation display module, which are used to integrate human decision-making information and realize electromagnetic spectrum situation visualization; the human-machine collaborative decision-making module uses prospect theory to quantify human cognitive bias and integrates human-side information and sensor data through OR / AND rules; the spectrum situation display module outputs electromagnetic spectrum distribution, signal location, and radiation source information in real time to support spectrum supervision and reconnaissance decision-making.

[0021] Therefore, the beneficial effects of the above-mentioned UAV-borne passive signal detection and positioning adaptive control method and system are as follows: (1) This invention constructs a full-link cognitive closed loop of “spectrum sensing - parameter decision - adaptive reception - identification direction finding and positioning - feedback optimization”, realizes dynamic adaptation of electromagnetic environment status and receiver operating parameters, solves the problem that traditional fixed parameter receivers cannot adapt to complex electromagnetic environment changes, and greatly improves the monitoring robustness of the system in complex background, strong clutter and strong interference environments. (2) By adopting a decision-making mechanism that combines convex optimization and reinforcement learning, the waveform parameters are optimized online adaptively. Combined with reconfigurable transceiver hardware, it can respond quickly to environmental changes and simultaneously take into account clutter suppression, interference suppression and small target detection performance. This solves the defects of existing technologies, such as difficulty in taking into account multiple targets and slow response speed. (3) By integrating a variety of advanced algorithms such as deep neural networks, attention mechanisms, space-time adaptive processing, and array direction finding and positioning, it has achieved accurate modeling of clutter interference, effective separation and identification of signals, and high-precision direction finding and positioning, which has significantly improved the accuracy of communication / radar signal monitoring, reconnaissance, and positioning in complex electromagnetic environments. (4) The system adopts an FPGA+DSP heterogeneous computing architecture and a reconfigurable hardware design to realize high-speed signal processing and real-time decision-making, which is suitable for the dynamic flight requirements of UAVs. At the same time, the system has a compact structure and low power consumption, which meets the load constraints of small UAVs. (5) The addition of a human-machine collaborative decision-making and spectrum situation display module improves the decision robustness in complex electromagnetic countermeasures environments, realizes electromagnetic spectrum situation visualization, expands the application scenarios of the system, and can be widely used in multiple fields such as radio monitoring.

[0022] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0023] Figure 1 This is a flowchart of an adaptive control method for passive signal reception and positioning on an unmanned aerial vehicle (UAV) according to the present invention. Figure 2 This is a block diagram of the structure of an unmanned aerial vehicle (UAV) passive signal reception, positioning, and adaptive control system according to the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0025] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0026] The following is combined Figures 1-2 The embodiments of the present invention will be described in detail below.

[0027] Example 1 This embodiment is based on a small unmanned aerial vehicle (UAV) platform, equipped with an electromagnetic spectrum monitoring receiver system, and applied to low-altitude spectrum monitoring scenarios to achieve the acquisition, identification, direction finding, and positioning of weak communication / radar signals in complex electromagnetic environments.

[0028] Step S1: Electromagnetic spectrum environment sensing and multi-source data fusion; The UAV's onboard spectrum monitoring receiver collects communication signals, radar signals, clutter signals (mainly ground clutter and meteorological clutter), and electromagnetic interference signals (mainly co-channel and adjacent-channel interference) in real time within the detection area (radius 5km, altitude 0-1km), at a sampling frequency of 100MHz. Simultaneously, the UAV's attitude angles (pitch, roll, and yaw) and flight speed (0-50m / s) are collected via the IMU module, and its position (latitude, longitude, and altitude) is collected via the GNSS module, constructing a multi-dimensional perception dataset.

[0029] A CNN-LSTM fusion neural network combined with an adversarial learning initialization strategy is employed to extract features from a perceptual dataset. The CNN layer uses a three-layer convolutional structure with kernel sizes of 3×3, 5×5, and 7×7 to extract local temporal and frequency domain features of the signal. The LSTM layer uses a two-layer bidirectional LSTM structure with 128 hidden units to capture the temporal correlation of the signal. The adversarial learning initialization strategy generates 10,000 sets of simulated low signal-to-noise ratio (SNR) signals, clutter, and interference samples using a GAN to pre-train the neural network.

[0030] The sensing dataset was classified and identified to determine clutter type (ground clutter, weather clutter), interference intensity (weak, medium, strong), signal type (communication / radar), and modulation method, thus achieving a preliminary understanding of the electromagnetic environment and potential target signals. Test results show that the accuracy of signal, clutter, and interference identification reaches 92%, meeting the system requirements.

[0031] Step S2: Receive parameters for adaptive decision optimization; A multi-objective optimization function is constructed with the optimization objectives of maximizing the weak signal detection probability, maximizing the interference suppression ratio, maximizing the direction finding accuracy, and minimizing the false alarm rate. The constraints are: receiver bandwidth ≤ 100MHz, gain ≤ 60dB, dynamic range ≥ 80dB, power consumption ≤ 15W, and flight attitude fluctuation ≤ 5°.

[0032] Combining the electromagnetic environment situation (mainly ground clutter, with moderate-intensity co-channel interference) output in step S1 with the preliminary signal characteristics (weak frequency-hopping communication signals, LFM radar signals), a convex optimization algorithm combining the MM algorithm and the Dinkelbach algorithm is adopted. Combined with the environment state-action-reward mapping of reinforcement learning, the optimal combination of receiving parameters is dynamically output.

[0033] Reinforcement learning employs the DeepQ-Network (DQN) algorithm, with a state space... S ={Clutter type, Interference level, Signal type, Signal-to-noise ratio}, Action space A ={frequency band switching, filter adjustment, gain adjustment, sampling rate adjustment, direction finding baseline configuration}, reward function R. After 5000 iterations of training, the algorithm converges to the optimal policy.

[0034] The optimal combination of receiving parameters for the final output is as follows: operating frequency band 2-6GHz, intermediate frequency bandwidth 20MHz, filter coefficient 0.8, receiving gain 45dB, sampling rate 100MHz, and direction finding baseline 1.2m.

[0035] Step S3: Adaptive signal reception and interference / clutter preprocessing; The reconfigurable filter amplifier completes frequency band switching, filter adjustment, and gain configuration within 80μs based on optimal receiving parameters. The configuration process includes parameter loading, filter generation, and gain amplification, with a total time of 80μs, meeting the system's real-time requirements. Electromagnetic signals are received through a multi-element receiving antenna.

[0036] The receiving end synchronously adjusts sampling parameters and employs STAP technology combined with a cognitive filtering algorithm to perform spatiotemporal joint preprocessing on the received signal. STAP uses a 3-element, 4-pulse spatiotemporal processing structure with 128 training samples. Covariance matrix estimation uses a diagonal loading technique with a loading amount of 10. -4 The cognitive filtering algorithm uses cosine similarity constraint for channel estimation, with a similarity threshold set at 0.85.

[0037] After preprocessing, the clutter suppression ratio reached 38dB and the interference suppression ratio reached 42dB, laying the foundation for subsequent signal identification and direction finding.

[0038] Step S4: Refined identification and direction finding of communication / radar signals; A ground clutter and meteorological clutter model was established using a stochastic transfer function combined with the Green's function impulse response principle. The model parameters were fitted using maximum likelihood estimation to achieve accurate clutter modeling. Test results show that the clutter modeling error is reduced by 35% compared to the traditional covariance statistical model.

[0039] A CNN algorithm with an attention mechanism is employed to separate communication signals, radar signals, clutter, and interference. The attention mechanism uses a multi-head self-attention structure with 8 heads and an attention dimension of 64. The CNN uses a residual network structure with a depth of 20 layers to extract the time-domain and frequency-domain features of the signal.

[0040] Based on the current clutter intensity (moderate) and interference intensity (moderate), the identification and direction finding thresholds are dynamically adjusted. A deep learning signal recognition + array interferometer direction finding + track fusion positioning algorithm is used to complete signal identification, direction finding, and positioning, outputting signal type, frequency, power, azimuth, and latitude / longitude location information.

[0041] Test results show that the signal recognition accuracy reaches 93%, the direction finding accuracy is ≤1°, the positioning error is ≤50m, and the false alarm rate is 8×10⁻⁶. -7 This meets the system design specifications.

[0042] Step S5: Closed-loop feedback iterative optimization; The identification effect, clutter suppression effect, direction finding accuracy, positioning error, and false alarm rate of step S4 are quantitatively evaluated. The evaluation indicators are: clutter suppression ratio = 38dB (≥35dB), interference suppression ratio = 42dB (≥40dB), signal identification accuracy = 93% (≥90%), direction finding accuracy = 0.8° (≤1°), positioning error = 42m (≤50m), and false alarm rate = 8×10⁻⁶. -7 (≤10) -6 ).

[0043] All evaluation indicators reached the preset thresholds. Since the electromagnetic environment remained largely unchanged, the closed-loop iteration period was set to 50ms to continuously monitor environmental changes. If the interference intensity subsequently increases, a new round of receiver parameter optimization will be initiated immediately.

[0044] Example 2 The system in this embodiment is used to implement the method in embodiment 1. The system is mounted on a small drone (weighing ≤5kg), and the specifications of each module are as follows: 1. Electromagnetic sensing module: It employs a high-speed RF acquisition chip (100MHz sampling rate, 16-bit), a multi-element receiving antenna (2-18GHz, gain ≥15dBi), an IMU module (measurement accuracy ±0.1°, sampling rate 100Hz), and a GNSS module (positioning accuracy ±1m). A CNN-LSTM fused neural network is deployed on an FPGA chip (Xilinx Zynq UltraScale + ZU7EV) to achieve feature extraction and recognition. FPGA resource configuration: 3604 DSP slices, 1824KB BRAM capacity, and 503200 logic units.

[0045] 2. Adaptive Decision Module: A heterogeneous computing architecture is adopted, consisting of an FPGA (Xilinx Zynq UltraScale+ZU7EV) and a DSP (TITMS320C6678). The FPGA is responsible for high-speed signal processing and rapid calculation of received parameters, while the DSP is responsible for multi-objective optimization algorithms and reinforcement learning policy execution. The FPGA and DSP are connected via an SRIO interface, with a data transmission rate of 5Gbps. The decision latency is 45μs, meeting real-time requirements.

[0046] The DSP has a clock speed of 1.25GHz, 8 cores, and a floating-point performance of 160GFLOPS. The reinforcement learning model is stored in the DSP's off-chip DDR4 memory (8GB capacity), and the model inference time is 30μs.

[0047] 3. Receiving Preprocessing Module: A multi-element interferometer receiving antenna (azimuth ±90°, elevation ±60°, beamwidth 0.5°-5°) is employed, with 8 elements and an operating frequency band of 2-18GHz. The reconfigurable filter amplifier supports 2-18GHz band switching, 10-100MHz bandwidth adjustment, and 0-60dB gain control, with a filter adjustment accuracy of 1MHz and a gain adjustment accuracy of 1dB. The high-speed sampler has a sampling rate of 100MHz and a bit depth of 16bit. The signal preprocessing module uses a high-speed filter chip (model ADIAD6688) to achieve initial suppression of clutter and interference.

[0048] 4. Identification and Direction Finding Positioning Module: The system deploys a CNN recognition algorithm with an attention mechanism, an array interferometer direction-finding algorithm, and a track fusion localization algorithm, employing a high-speed signal processing chip (Xilinx Zynq UltraScale+ZU9EG) to achieve signal separation and recognition, direction finding, and localization. Chip resource configuration: 2520 DSP slices, 1296KB BRAM, and 460800 logic units. Processing latency ≤30μs.

[0049] 5. Closed-loop feedback module: A performance evaluation unit and a feedback control unit are employed to collect recognition results, direction finding and positioning accuracy, and suppression effects in real time. These results are then quantitatively evaluated and fed back to the adaptive decision-making module. The performance evaluation unit uses hardware counters to statistically analyze indicators such as recognition accuracy, direction finding accuracy, and false alarm rate. The feedback control unit uses a state machine to control the iterative process. The iteration period can be dynamically adjusted within the range of 10ms-100ms.

[0050] 6. Data storage module: It uses a 128GB high-speed solid-state drive (Samsung 970EVOPlus) with a data transfer rate of 1.2Gbps to store sensing data, historical received parameters, identification results, direction finding and positioning data, and system logs. The file system uses the ext4 format, supporting real-time data writing and reading.

[0051] 7. Human-machine collaborative decision-making module and spectrum situation display module: The human-machine collaborative decision-making module supports manual intervention to adjust parameters, which is achieved through a ground control station. The spectrum situation display module outputs electromagnetic spectrum distribution, signal location, and radiation source information in real time, supporting spectrum monitoring and reconnaissance decision-making.

[0052] The system in this embodiment operates in low-altitude spectrum monitoring scenarios, effectively suppressing ground clutter, meteorological clutter, and co-channel and adjacent-channel interference. It accurately captures, identifies, locates, and positions weak communication / radar signals, exhibiting high monitoring accuracy and robustness. It is suitable for the dynamic flight requirements of UAVs and can operate stably in complex electromagnetic environments.

[0053] The system in this embodiment operates in low-altitude security scenarios, effectively suppressing ground clutter, weather clutter, and suppression interference. It accurately detects small targets such as small drones and birds, exhibiting high monitoring accuracy and robustness. It is adapted to the dynamic flight requirements of drones and can operate stably in complex environments.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive control method for passive signal reception and positioning on an unmanned aerial vehicle (UAV), characterized in that, Includes the following steps: S1. The UAV-borne cognitive radar collects communication signals, radar signals, clutter signals, and electromagnetic interference signals in the detection area in real time, and simultaneously integrates attitude, position, and velocity information output by the UAV's inertial measurement unit (IMU) and global navigation satellite system (GNSS) module to construct a multi-dimensional perception dataset; based on the perception dataset, electromagnetic environment and signal features are extracted, and clutter type, interference intensity, interference level, signal type, and modulation method are identified to complete the preliminary understanding of the electromagnetic environment situation and potential target signals; S2. A multi-objective optimization function is constructed with the optimization objectives of maximizing the weak signal detection probability, maximizing the interference suppression ratio, maximizing the direction finding accuracy, and minimizing the false alarm rate. Combining the electromagnetic environment situation, preliminary signal characteristics, and the receiver's historical operating status, the optimal combination of receiving parameters is dynamically output through the optimization algorithm. S3. Based on the optimal receiving parameters, drive the receiver front-end to complete frequency band adaptive switching, real-time adjustment of filter parameters, adaptive control of receiving gain, and synchronously adjust multi-channel sampling and array preprocessing parameters; perform space-time joint preprocessing on the received signal to suppress clutter and electromagnetic interference. S4. Perform precise clutter interference modeling and signal separation on the preprocessed signal from S3, and extract communication / radar signal features; dynamically adjust the identification and direction finding thresholds based on the current clutter and interference intensity to complete communication / radar signal identification, high-precision direction finding and radiation source localization under complex backgrounds, and output target information. S5. Quantitatively evaluate the signal recognition effect, clutter interference suppression effect, direction finding accuracy, positioning error and false alarm rate; feed the evaluation results back to step S2, iteratively optimize the receiving parameters, and form a cognitive closed loop of "spectrum sensing - parameter decision - adaptive reception - identification, direction finding and positioning - feedback optimization".

2. The adaptive control method for UAV-borne passive signal reception and positioning according to claim 1, characterized in that, The multi-dimensional sensing dataset in S1 includes the time-domain characteristics, frequency-domain characteristics, polarization characteristics, and modulation characteristics of communication / radar signals; the amplitude distribution and power spectrum characteristics of clutter; the frequency, power, and modulation characteristics of electromagnetic interference; and the attitude angle, flight speed, and altitude information of UAVs. Clutter types include ground clutter, sea clutter, and meteorological clutter; and interference types include co-channel interference, adjacent-channel interference, pulse interference, and suppression interference. Signal types include communication signals and radar signals.

3. The adaptive control method for UAV-borne passive signal reception and positioning according to claim 2, characterized in that, S1 employs a deep neural network for feature extraction and classification, and uses an adversarial learning initialization strategy to improve the recognition accuracy of signals, clutter, and interference under low signal-to-noise ratio conditions. The deep neural network adopts a fusion architecture of convolutional neural network (CNN) and long short-term memory network (LSTM). CNN is used to extract local features in the time and frequency domains of the signal, while LSTM is used to capture the temporal correlation of the signal. The adversarial learning initialization strategy generates simulated signal, clutter, and interference samples under low signal-to-noise ratio conditions through generative adversarial network (GAN) to pre-train the neural network.

4. The adaptive control method for UAV-borne passive signal reception and positioning according to claim 3, characterized in that, The specific details of S2 are as follows: S21 Construct a multi-objective optimization function The expression is as follows: ； In the formula, For weak signal detection performance indicators; For interference suppression performance indicators; For direction finding accuracy indicators; The false alarm rate is an indicator. To receive the parameter vector; satisfying ; The constraints of the multi-objective optimization function include bandwidth limitations, gain limitations, dynamic range limitations, power consumption limitations, and flight attitude constraints of the UAV-borne receiver. The receiver parameter combination includes operating frequency band, intermediate frequency bandwidth, filter coefficient, receiver gain, sampling rate, and direction finding baseline configuration; it supports multi-band adaptive switching and dynamic adjustment of multiple filter modes; S22 employs a decision-making mechanism combining convex optimization and reinforcement learning, transforming a non-convex optimization problem into a fast-solvable convex optimization problem. It also achieves online adaptive adjustment of waveform parameters through an environment state-action-reward mapping, as shown in the following expression: ； In the formula, The transformed convex optimization objective function; This is the negative form of the original objective function; The regularization coefficient is used. It is a regular term; The method employs an iterative solution combining the MM algorithm and the Dinkelbach algorithm, and constructs an environment state-action-reward mapping through reinforcement learning; the receiving parameters are then adaptively adjusted online through this environment state-action-reward mapping.

5. The adaptive control method for UAV-borne passive signal reception and positioning according to claim 4, characterized in that, In S3, frequency band switching, filtering adjustment, and gain control are performed through an adaptive RF front-end; the receiver uses space-time adaptive processing STAP combined with a cognitive filtering algorithm to perform preliminary suppression of clutter and interference; the cognitive filtering algorithm, combined with a channel estimation method based on cosine similarity constraints, reduces filtering errors under low signal-to-noise ratio; the STAP technology estimates the covariance matrix through training basis to shape an adaptive receiving beam.

6. The adaptive control method for UAV-borne passive signal reception and positioning according to claim 5, characterized in that, In S4, a stochastic transfer function is used to establish a clutter interference model, and a deep learning algorithm with an attention mechanism is combined to separate communication signals, radar signals, clutter, and interference. The stochastic transfer function, based on the Green's function impulse response principle, performs accurate clutter modeling, and its expression is as follows: ； In the formula, This is a clutter signal; It is the Green's function impulse response; For random excitation sources; To observe noise; Signal recognition, direction finding, and localization are achieved by employing deep learning signal recognition, array interferometer / spatial spectrum estimation direction finding, and multi-station time difference / frequency difference positioning technology. The detection and recognition threshold is adjusted in real time according to the complexity of the environment.

7. The adaptive control method for UAV-borne passive signal reception and positioning according to claim 6, characterized in that, The quantitative evaluation indicators in S5 include clutter suppression ratio, interference suppression ratio, signal recognition accuracy, direction finding accuracy, positioning error, and false alarm rate; preset performance thresholds for clutter suppression ratio, interference suppression ratio, signal recognition accuracy, direction finding accuracy, positioning error, and false alarm rate; when the evaluation indicators do not reach the preset thresholds, a new round of receiver parameter optimization is immediately initiated, and the closed-loop iteration cycle is dynamically adjusted according to the complexity of the environment.

8. An adaptive control system for passive signal reception, positioning, and control on an unmanned aerial vehicle (UAV), characterized in that, The UAV-borne passive signal detection and positioning adaptive control method according to any one of claims 1-7 includes: an electromagnetic sensing module, an adaptive decision module, a receiving preprocessing module, an identification direction finding and positioning module, a closed-loop feedback module, and a data storage module; each module realizes real-time data interaction through a high-speed data bus; The electromagnetic sensing module is used to receive communication / radar signals, clutter, interference signals, and UAV status information, and to extract electromagnetic environment and signal features. Specifically, it includes a multi-channel receiving antenna unit, a radio frequency acquisition unit, a data fusion unit, and a feature extraction and recognition unit. The multi-channel receiving antenna unit is used to receive electromagnetic signals, the radio frequency acquisition unit is used for high-speed sampling, the data fusion unit is used to fuse the status information output by the UAV's IMU and GNSS modules to construct a multi-dimensional sensing dataset, and the feature extraction and recognition unit uses a deep neural network to complete feature extraction and signal, clutter, and interference recognition. The adaptive decision-making module is used to construct a multi-objective optimization function and dynamically decide the optimal receiving parameters. It includes a feature recognition unit, a multi-objective optimization unit, and a reinforcement learning unit. The feature recognition unit uses a deep neural network model to extract and recognize features from a multi-dimensional sensing dataset and outputs preliminary characteristics of the electromagnetic environment and signals. The multi-objective optimization unit constructs a multi-objective optimization function and solves for the optimal combination of receiving parameters through a convex optimization algorithm. The reinforcement learning unit combines historical data and real-time feedback information to optimize the receiving parameter decision-making strategy. The receiving preprocessing module is used to perform adaptive reception, beamforming, and signal preprocessing based on the receiving parameters. It includes an adaptive radio frequency unit and a cognitive preprocessing unit. The adaptive radio frequency unit includes a reconfigurable filter amplifier and a high-speed sampler, which are used to achieve adaptive frequency band, filtering, gain configuration, and signal acquisition based on the optimal receiving parameters output by the optimization decision module. The cognitive preprocessing unit includes a signal preprocessing module, which is used to acquire electromagnetic signals, perform preliminary filtering and preprocessing, and then transmit them to the identification, direction finding, and positioning module. The identification, direction finding, and positioning module is used for signal separation, communication / radar signal identification, high-precision direction finding, and radiation source positioning. It employs technologies such as accurate clutter interference modeling, deep learning separation and identification, array direction finding, and track fusion positioning to refine the pre-processed signal, extract signal features, and complete accurate identification of communication / radar signals, estimation of incoming wave direction, and radiation source positioning. It outputs signal type, frequency, power, azimuth angle, and latitude and longitude location information. The closed-loop feedback module is used for performance evaluation and feedback iteration to form a cognitive closed loop; it collects the identification results, direction finding and positioning accuracy, clutter and interference suppression effect of the identification and direction finding and positioning module, constructs a performance evaluation index system, evaluates the adaptability of the current received parameters, transmits the evaluation feedback information to the optimization decision module, and initiates the parameter iteration optimization process. The data storage module is used to store sensing data, historical parameters, recognition results, direction finding and positioning data, and system logs; it adopts a high-speed storage chip to support real-time writing and reading of data, providing data support for optimization decisions and iterative upgrades.

9. The UAV-borne passive signal reception, positioning, and adaptive control system according to claim 8, characterized in that: The receiver preprocessing module uses a multi-element interferometer / phased array receiving antenna, which supports adaptive receiving beamforming and enables dynamic adjustment of receiving direction and range; the reconfigurable filter amplifier supports real-time switching of multiple frequency bands, multiple filtering modes, and multiple gain levels. The adaptive decision-making module adopts a heterogeneous computing architecture of FPGA+DSP. The FPGA is used to realize high-speed signal processing and fast calculation of received parameters, while the DSP is used to implement complex optimization algorithms and reinforcement learning strategies. The heterogeneous computing architecture realizes parallel computing of feature recognition, multi-objective optimization and reinforcement learning, ensuring the real-time performance of received parameter optimization decisions.

10. The UAV-borne passive signal reception, positioning, and adaptive control system according to claim 9, characterized in that: The system also includes a human-machine collaborative decision-making module and a spectrum situation display module, which are used to integrate human decision-making information and realize electromagnetic spectrum situation visualization. The human-machine collaborative decision-making module uses prospect theory to quantify human cognitive bias and integrates human-side information and sensor data through OR / AND rules. The spectrum situation display module outputs electromagnetic spectrum distribution, signal location, and radiation source information in real time to support spectrum supervision and reconnaissance decision-making.

Images