A low-altitude-oriented unmanned aerial vehicle perception method

Abstract

The application discloses a low-altitude-oriented unmanned aerial vehicle sensing method and relates to the technical field of unmanned aerial vehicle sensing. The method utilizes 5G downlink signals, decodes and re-encodes reference sending data, and simultaneously aligns the direction of the unmanned aerial vehicle with the aid of array antenna beamforming. Element-by-element ratio operation is performed on the received signals and the reference data to eliminate the influence of random data, thereby reconstructing the channel state information of the reflection path, generating a distance-Doppler graph with high signal-to-noise ratio, and realizing unmanned aerial vehicle detection and parameter estimation. The method converts data symbols into sensing resources without changing the existing 5G waveform, breaks through the sensing range limitation caused by the sparsity of reference signals in traditional methods, and significantly improves the detection distance and the stability and accuracy of low-altitude unmanned aerial vehicle sensing.

Description

Technical Field

[0001] This invention relates to the field of drone sensing technology, and more particularly to a drone sensing method for low-altitude operations. Background Technology

[0002] With the rapid proliferation of drones, safety issues in low-altitude airspace have become increasingly critical. Multiple drone operations have led to mid-air collisions and accidents, and unauthorized incursions into restricted areas such as airports have become more common. To address these issues, existing technologies primarily employ visual, acoustic, and radar-based solutions.

[0003] However, existing drone sensing solutions face inherent limitations: visual and acoustic systems are susceptible to environmental interference, while radar solutions typically require high hardware costs. Furthermore, existing approaches all require dedicated infrastructure, limiting their scalability and cost-effectiveness in large-scale deployments, and presenting technical challenges such as limited range or high hardware costs. The dense deployment of 5G infrastructure offers new opportunities for drone sensing, with its high bandwidth and massive MIMO technology providing high-resolution sensing capabilities.

[0004] However, 5G waveforms are designed with communication at their core, which leads to insufficient perceived signal-to-noise ratio (SSNR) for channel state information (CSI) estimation, fundamentally limiting the sensing range. Currently, the maximum detection range of existing technologies is only about 250 meters. In addition, existing 5G-based sensing schemes often use waveform modification or only utilize sparse reference signals, reducing communication capabilities and failing to achieve high-performance sensing over long distances. Summary of the Invention

[0005] This application provides a low-altitude drone sensing method that solves the problems of insufficient sensing signal-to-noise ratio and limited detection range caused by the scarcity of reference signal resources in the prior art. It realizes the technical effect of converting data payload symbols in 5G signals into sensing resources, thereby significantly improving the sensing range.

[0006] This application provides a perception method for low-altitude unmanned aerial vehicles (UAVs), including: S1: Receive 5G downlink signal, decode the downlink signal to obtain decoded bit stream, and then re-encode the decoded bit stream to recover the corresponding time-frequency resource grid as reference data; at the same time, according to the antenna array of the receiving device, form a receiving beam to enhance the reflection path of the target UAV through beamforming; S2: Perform an element-wise ratio operation on the time-frequency resource cell corresponding to the received beam output signal and the time-frequency resource cell corresponding to the recovered reference transmitted data to eliminate the influence of random data load and reconstruct the channel state information of the target reflection path; perform a two-dimensional Fourier transform on the channel state information to generate a range-Doppler map; S3: Based on the range-Doppler graph, detect and estimate the distance and radial velocity of the target UAV.

[0007] The channel state information includes all downlink resource elements and corresponds one-to-one with the target reflection path.

[0008] Furthermore, the 5G downlink signal is decoded to restore the time-frequency resource format consistent with the transmitter, including: completing downlink synchronization, orthogonal frequency division multiplexing demodulation, physical channel demodulation and channel decoding to obtain the downlink data bit stream; and according to the 5G physical layer mapping rules, the data bit stream is remodulated and resource mapped to restore the time-frequency resource format consistent with the transmitter.

[0009] Furthermore, beamforming is used to form a receiving beam to enhance the reflection path of the target UAV, including: predefining an azimuth angle range, generating several candidate beamforming weight vectors to form multiple candidate beams; for each candidate beam, performing element-wise division calculation of the time-frequency resource grid and channel state information reconstruction, and calculating the perceived signal-to-noise ratio of the target peak in the range-Doppler image, and selecting the candidate beam with the highest perceived signal-to-noise ratio as the receiving beam; the perceived signal-to-noise ratio is defined as the ratio of the target peak power to the background noise power.

[0010] Furthermore, the process of detecting and estimating target parameters based on the range-Doppler image includes: firstly, detecting potential target peaks in the image; performing trajectory matching using a temporal correlation algorithm on target peaks detected in multiple consecutive time frames to filter out false detections; and outputting estimated values ​​of distance and velocity for successfully correlated target trajectories. The distance estimate is determined by the range cell index where the target peak is located, the speed of light, the total number of subcarriers, and the subcarrier spacing. The velocity estimate is determined by the Doppler cell index where the target peak is located, the carrier wavelength, the total number of symbol frames, and the duration of each symbol.

[0011] Furthermore, a Fourier transform is performed on the channel state information matrix along the subcarrier dimension to convert it into a range profile, where paths with different time delays will form peaks in different range cells. Simultaneously, a Fourier transform is performed along the continuous symbol time dimension to convert it into a Doppler profile, where Doppler frequency shifts caused by different velocities will form peaks in different Doppler cells. Combining these two one-dimensional transforms into a two-dimensional Fourier transform yields the final range-Doppler map, where the transform kernel function is responsible for mapping the data from the channel domain to the range-velocity domain.

[0012] Further, generating the beamforming weight vector includes: beam guiding based on the antenna array, assuming the signal is incident from a specific angle, for two adjacent antenna elements, the path difference will cause a phase difference in the received signal, which is determined by the antenna spacing, incident angle and carrier wavelength; by constructing a guiding vector whose elements consist of a series of complex weights to compensate for the aforementioned phase difference, and multiplying the signal received by each antenna by the corresponding weight and then merging them, the signal from that specific direction can be enhanced by in-phase superposition, while the signal from other directions is suppressed.

[0013] One or more technical solutions provided in this application have at least the following technical effects or advantages: The fundamental bottleneck of existing 5G-based drone sensing technology lies in the severe shortage of sensing resources. 5G waveforms are designed with communication at their core, and the reference signal used for channel estimation typically accounts for less than 10% of the time-frequency resources to ensure spectral efficiency. This results in an excessively small number of effective resource elements, a key parameter in the signal-to-noise ratio (SNR) of reference signal-based sensing. Experiments show that when the target distance increases to approximately 250 meters, the SNR drops below the reliable detection threshold, leading to sensing failure. Existing improvement schemes either attempt to modify the standard waveform (affecting communication) or can only utilize sparse reference signals, failing to overcome this resource limitation and thus restricting their maximum detection range and practicality.

[0014] Based on an understanding of the aforementioned problems, this invention breaks away from the conventional thinking of relying solely on known reference signals and creatively proposes a novel path to transform approximately 90% of the unknown data payload symbols in 5G downlink frames into sensing resources. The core idea is to utilize the discrete characteristics of modulation symbols in symbol space, reliably recover the modulation symbols through decoding operations, reconstruct the complete time-frequency resource grid, and use it as a reference for data transmission, thereby unlocking all downlink resources for sensing.

[0015] By calculating the element-wise ratio of the time-frequency resource grid corresponding to the received beam output signal to the reconstructed complete time-frequency resource grid, the complete channel state information of the target path can be reconstructed. This method is effective for all time-frequency resource elements (including data symbols), increasing the number of effective sensing resources by nearly an order of magnitude. Based on the sensing signal-to-noise ratio model, this directly translates to a significant gain. Field experiments confirm that, under the same low transmit power, the method in this application significantly extends the maximum reliable detection range of UAVs from 245 meters using traditional methods to 416 meters, an improvement of 69.8%. Simultaneously, it maintains sub-meter range accuracy and a speed accuracy better than 0.5 meters per second across the entire range from 0 to 400 meters. Furthermore, it demonstrates robustness in non-line-of-sight and multi-target scenarios, providing a practical and optimal solution for large-scale, low-cost, and high-performance low-altitude UAV surveillance. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a low-altitude drone perception method according to an embodiment of the present invention. Detailed Implementation

[0017] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0019] To verify the new understanding of the core defect of existing 5G sensing technology—that the sparse reference signal leads to a limited sensing range—and to confirm the effectiveness of the proposed new path of using decoded and reconstructed time-frequency resource grids as references for data transmission and unlocking data payload symbols for sensing, this embodiment aims to specifically construct and evaluate the core solution of the present invention.

[0020] The technical terms involved in this invention are explained as follows: Channel State Information (CSI) is a term used in the field of wireless communication. It refers to the channel attributes of a communication link and is used to describe the attenuation factor of a signal on each transmission path, that is, the value of each element in the channel gain matrix.

[0021] RE stands for Resource Element, which is the smallest physical resource unit in a wireless communication system. In the time domain, it corresponds to an OFDM symbol, and in the frequency domain, it corresponds to a subcarrier.

[0022] OFDM (Orthogonal Frequency Division Multiplexing) is a type of MCM (Multi Carrier Modulation) that uses frequency division multiplexing to achieve parallel transmission of high-speed serial data.

[0023] SSNR stands for Signal-to-Noise Ratio, which is the ratio of the strength of the received useful signal to the strength of the received interference signal.

[0024] In this embodiment, a software-defined radio-based 5G transmitting base station (using the OAI5GRAN software stack) needs to be pre-deployed to continuously transmit standard-compliant 5G downlink waveforms, including sparse reference signals and randomly generated data payloads. A receiving base station equipped with a four-antenna array is deployed to receive multipath signals including direct path, environmental reflections, and drone reflections. The transmit power is set to 11dBm (12.6mW), operating in the 3.3GHz band with a bandwidth of 40MHz, using a commercial drone as the target.

[0025] 5G employs Orthogonal Frequency Division Multiplexing (OFDM) modulation, and 5G transmissions are organized into 10-millisecond radio frames in the time domain. A radio frame is divided into 1-millisecond subframes, each containing one or more time slots, depending on the subcarrier spacing. After transmission from the base station, the 5G signal propagates through multiple radio paths to reach the receiver. These radio paths include not only reflections from drones but also non-target components such as LOS paths and environmental reflections. 5G defines two types of physical resources: physical channels and reference signals. Physical channels carry dynamic payloads, including control information and user data. In contrast, reference signals are predefined sequences used to facilitate channel estimation and synchronization. Common types include DMRS, phase tracking reference signals, CSI-RS, and tracking reference signals. Data payload symbols occupy the majority of REs to maintain high communication throughput, while reference signals are sparsely distributed, occupying only a small portion of the resources.

[0026] The receiving base station performs frame synchronization on the received time-domain signal by cross-correlating the received samples with a locally generated synchronization signal block (SSB) template and detecting peak values ​​to determine the frame start position. After synchronization, the original signal is divided into 10ms radio frames, and carrier modulation and demodulation are performed on each frame: the cyclic prefix is ​​removed and a Fourier transform is applied to each carrier modulation symbol, thereby converting the original signal into a two-dimensional time-frequency resource grid, where each resource element corresponds to a specific subcarrier and carrier modulation symbol.

[0027] Example 1: As Figure 1 As shown, a low-altitude drone sensing method includes: receiving a 5G downlink signal, decoding the downlink signal to obtain a decoded bitstream, and then re-encoding the decoded bitstream to recover the corresponding time-frequency resource grid as reference data; simultaneously, based on the antenna array of the receiving device, forming a receiving beam to enhance the reflection path of the target drone through beamforming.

[0028] In this embodiment, forming a time-frequency resource grid corresponding to the reference transmitted data includes: synchronizing the received 5G downlink signal, performing orthogonal frequency division multiplexing demodulation, physical channel demodulation, and channel decoding to obtain a downlink data bit stream; based on the data bit stream, remodulating the data bit stream according to the 5G physical layer modulation and resource mapping rules and mapping it to the corresponding time-frequency resource location, thereby restoring a time-frequency resource grid consistent with the transmitter.

[0029] In some embodiments, forming a receiving beam includes: predefining an azimuth angle range, generating several candidate beamforming weight vectors to form multiple candidate beams; within a preset azimuth angle range, for example, within a ±60° sector area in front of the receiving base station, selecting a series of candidate angles at certain intervals (e.g., 1°), and sequentially substituting them into the above formula to calculate the corresponding vectors, generating a beamforming signal and weighting the received signal to generate multiple candidate beamforming weight vectors. For each candidate beam, performing element-wise division of the time-frequency resource grid and channel state information reconstruction, and calculating the perceived signal-to-noise ratio of the target peak in the obtained range-Doppler image; selecting the candidate beam with the highest perceived signal-to-noise ratio as the receiving beam.

[0030] In some embodiments, in conventional CSI-based sensing, the sensing signal-to-noise ratio (SNR) metric is typically used to quantify sensing quality; it measures the power ratio between dynamic and static CSI components. In 5G drone sensing, parameter estimation depends only on the target peak value in the range-Doppler map. Therefore, in this application, an SSNR model is constructed, whereby the sensing SNR is defined as: ,in, and These represent the power of the target peak and the power of the background noise on the range-Doppler map, respectively. The candidate beam with the highest perceived signal-to-noise ratio (SSNR) is selected as the receiving beam. A higher SSNR indicates a more significant distinction between the target peak and the noise floor, leading to more stable detection. Detection performance at different SSNR levels is experimentally verified, showing robust performance at high (30.2 dB) and medium (16.5 dB) SSNR. When the SSNR drops to 6.7 dB, the target peak is submerged in the noise floor, resulting in detection failure. Observations confirm that the proposed SSNR metric effectively characterizes perception quality in the range-Doppler domain. The SSNR in the range-Doppler map is proportional to the total number of reference signals in the physical 5G frame but inversely proportional to the fourth power of the UAV distance; this means that increasing the resource elements in the channel state information matrix can mitigate the SSNR loss caused by distance, thereby extending the effective sensing range.

[0031] In practical experiments, this application validates the proposed SSNR model using sensing data collected from a USRPB210 base station. An ideal downlink signal is generated using the MATLAB 5G toolbox, allowing for flexible control of the reference signal ratio. A commercial drone is used as the sensing target. Range-Doppler plots are generated under different reference signal densities. Experimental verification shows that as more REs are allocated to the reference signal, the target correlation peak at 296 meters becomes increasingly prominent. This trend is further quantified by plotting the relationship between measured SSNR and reference signal density. At a 10% ratio, the SSNR is only 7.39 dB. As the ratio increases, the SSNR rises sharply, reaching 17.66 dB when all REs are used as the reference signal. The observed trend closely matches the predictions of the theoretical SSNR model in this application.

[0032] In this application, to enhance the SSNR of UAV detection, the number of resource elements available for channel state information estimation must be maximized. Besides the dedicated reference signal, the 5G downlink frame contains a large number of data payload symbols. Experimental verification shows that in actual 5G downlink frames captured from the OAI5G protocol stack, the reference signal is sparsely distributed on the time-frequency grid. In contrast, the data payload symbols are densely arranged, constituting nearly 90% of all downlink resource elements. Since both symbol types traverse the same radio channel... The data payload symbols theoretically provide additional observations that can be incorporated into the CSI estimate. Utilizing these symbols will enable the receiver to take advantage of the full set of REs, thus providing a significant SSNR improvement.

[0033] However, utilizing data payloads for sensing presents significant challenges. Reference signals in 5G frames use predefined sequence coding, allowing for direct CSI estimation via simple division. In contrast, data payload symbols carry dynamic user information with randomized content, preventing the direct application of traditional division-based channel estimation methods to data symbols.

[0034] Based on the above reasons, the method provided in this application selects to use the standard 5G decoding process to obtain the decoded bit stream and reconstruct the time-frequency resource grid corresponding to the transmitted data, which actually utilizes the discrete characteristics of the modulation symbols in the symbol space.

[0035] In some embodiments, generating a beamforming weight vector includes: performing beam control based on the MIMO antenna array, and presetting the angle of arrival as... For the distance between Two adjacent antennas, the phase shifts of the two incident signals are respectively ,in, It is the carrier wavelength; the steering vector received by the base station, as the beamforming weight vector, is: ,in, This represents the total number of antenna elements in the antenna array; phase weighting compensates for the spatial phase across the array by multiplying the received signal at each antenna by the corresponding weight in the steering vector, aligning the phase of all elements with the phase of the first antenna. The aligned signals are then coherently summed, resulting in a phase from the direction... The energy-constructive properties of the signal are enhanced. In contrast, signals from other directions have mismatched phases, thus combining destructively during summation, resulting in suppressed output. A receiving beam (i.e., a candidate beam) is formed based on spatial filtering effects, selectively enhancing the signal from the UAV while suppressing interference from other angles.

[0036] In specific experiments, this application, based on a beamforming model, considers the application of beamforming to enhance signals from a UAV in UAV perception scenarios with both target and non-target signal paths. Since the signal paths originate from different incident angles, to enhance the UAV's reflection component, the base station first estimates the UAV's direction, denoted as... Then, a beam control vector is applied in that direction, and the resulting received beam output signal is: , It is the drone's reflection path, determined by time delay. and Doppler shift Characterization; These represent the attenuation coefficient, subcarrier spacing, and carrier modulation symbol duration of the UAV's reflection path, respectively. , It is a complete transmission symbol, including the known reference signal. and data payload ; This represents additive white Gaussian noise in the received signal.

[0037]

[0038] in, It is the transmission power; and These are the antenna gains of the transmitting base station and the receiving base station, respectively. It refers to the carrier wavelength, specifically the carrier wavelength of the 5G downlink signal in the air in this application; It is the radar cross-section of the drone; It is the actual distance of the drone's reflected path.

[0039] The channel state information is reconstructed by performing an element-by-element ratio calculation between the time-frequency resource cell corresponding to the received beam output signal and the time-frequency resource cell corresponding to the recovered reference transmitted data. ,in, It is the time-frequency resource grid that has been restored after decoding. It is the time-frequency resource grid corresponding to the received beam output signal. It is the reconstructed channel state information.

[0040] A two-dimensional Fourier transform is performed on the reconstructed channel state information to generate a range-Doppler map. Generating the range-Doppler map includes: performing a range-Fourier transform across subcarriers to convert the frequency-domain channel state information matrix into a range profile, where each path generates a different peak at a specific range bin; performing a Doppler Fourier transform across consecutive carrier modulation symbols to convert the phase change caused by the Doppler frequency shift into a Doppler profile; combining these two transforms into a joint two-dimensional Fourier transform to obtain the range-Doppler map, as follows: ,in, It is to transform the channel state information matrix to the first... The distance warehouse and the first Fourier transform of a Doppler chamber; This represents the total number of subcarriers. is the total number of carrier modulation symbol frames; k is the subcarrier index, representing the kth subcarrier; s is the carrier modulation symbol frame index, representing the sth carrier modulation symbol frame.

[0041] Based on the range-Doppler map generated from the reconstructed channel state information, the distance and radial velocity of the target UAV are detected and estimated. The process of detecting and estimating the distance and radial velocity of the target UAV based on the range-Doppler map includes: detecting potential target peaks in the range-Doppler map using a two-dimensional constant false alarm rate (CFAR) detector; performing trajectory association using a temporal correlation algorithm on the target peaks detected in multiple consecutive frames to filter out false detections; and outputting estimated UAV distance and velocity values ​​for the associated target trajectory. The UAV distance is: The drone's speed is: ;in, and These are the range cell and Doppler cell indices corresponding to the target peak in the range-Doppler map, respectively; for Speed ​​of light; Subcarrier spacing; The carrier wavelength; The duration of the modulation symbol for the s-th carrier; This represents the total number of subcarriers. This represents the total number of carrier modulation symbol frames.

[0042] This application uses an airborne experiment to verify the proposed channel estimation method. A USRPB210 running the OAI5G protocol stack is used as the transmitter, continuously transmitting downlink frames containing a reference signal and randomly generated payload data. At the receiver, a USRPX310 equipped with a four-antenna array implements the proposed channel estimation based on decoded and reconstructed time-frequency resource grids as references for transmitting data.

[0043] The CSI obtained by the two methods was compared. Since the 5G reference signal only accounts for a small fraction of REs, the conventional estimator generates CSI on a sparse set of subcarriers. In contrast, the decode-reconstructed reference transmission data method reconstructs CSI on all subcarriers, producing a significantly denser channel profile. CSI based on decode-reconstructed reference transmission data achieves a significantly higher SSNR, and the UAV remains clearly detectable even at a distance of 380m. These results demonstrate that the proposed channel estimation method effectively improves SSNR and significantly extends the UAV's sensing range.

[0044] This embodiment fully demonstrates the entire process from signal reception to parameter estimation. Its core effect is that it clarifies the bottleneck of traditional methods through the SSNR model; by decoding and reconstructing the time-frequency resource grid as a reference for transmitting data, the energy of the data payload symbols is incorporated into the sensing, which increases the effective resource number by nearly an order of magnitude and directly translates into a significant gain in SSNR, laying a theoretical foundation for breaking through distance limitations.

[0045] The technical solution proposed in this application enables long-range monitoring of the distance and speed of drones. Results show that, with a transmission power of only 11 dBm, it can accurately detect drones at a distance of 416 meters, expanding the sensing range by 80% compared to traditional methods while maintaining high accuracy in distance and speed estimation.

[0046] This application receives 5G downlink signals, decodes them to obtain a decoded bitstream, and then re-encodes the decoded bitstream to recover the corresponding time-frequency resource grid, which is then used as a reference for data transmission.

[0047] Within a preset azimuth range, multiple candidate beams (receive beam candidates) are generated through scanning to cover potential UAV arrival directions. For each candidate beam, element-wise division of the time-frequency resource grid and channel state information reconstruction are performed, and the perceived signal-to-noise ratio (SNR) of the target peak in the range-Doppler graph is calculated. The candidate beam with the highest perceived SNR is selected as the receive beam, and beamforming is used to further improve the SSNR.

[0048] By dividing the time-frequency resource grid corresponding to the received beam output signal element-wise with the time-frequency resource grid corresponding to the recovered reference transmitted data, dense channel state information of the target path is reconstructed. A two-dimensional Fourier transform is then performed to generate a high signal-to-noise ratio range-Doppler map. Finally, through constant false alarm rate detection, trajectory correlation, and filtering algorithms, the accurate range and radial velocity of the UAV are stably output.

[0049] This embodiment directly solves the core problems of low sensing signal-to-noise ratio and short effective detection distance caused by sparse reference signals in standard 5G sensing; at the same time, it solves the problem of how to securely and effectively use downlink resource symbols, which account for a very large proportion and carry unknown user data, for radar-like sensing without affecting communication.

[0050] Through the systematic integration of the above-mentioned technical means, this application achieves the following effects: By increasing the number of effective sensing resources, the stable and reliable detection range of UAVs has been significantly increased from approximately 245 meters using traditional methods to 416 meters at extremely low transmission power, representing an absolute increase of 171 meters and a relative increase of 69.8%.

[0051] While greatly expanding the detection range, it successfully maintained sub-meter-level distance estimation accuracy and velocity estimation accuracy better than 0.5 meters per second throughout the new range, meeting the accuracy requirements of practical regulatory applications.

[0052] The feasibility of the solution in a real-world environment was verified, using a standard open-source 5G protocol stack and a general-purpose software radio device; signal generation simulated real communication traffic.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A sensing method for low-altitude unmanned aerial vehicles (UAVs), characterized in that, include: S1: Receive 5G downlink signal, decode 5G downlink signal to obtain decoded bit stream, and then re-encode the decoded bit stream to recover the corresponding time-frequency resource grid as reference data; at the same time, according to the antenna array of the receiving device, form a receiving beam to enhance the reflection path of the target UAV through beamforming; S2: Perform an element-wise ratio operation between the time-frequency resource cell corresponding to the received beam output signal and the time-frequency resource cell corresponding to the recovered reference transmission data to reconstruct the channel state information of the target reflection path; perform a two-dimensional Fourier transform on the channel state information to generate a range-Doppler map; S3: Based on the range-Doppler map, detect and estimate the range and radial velocity of the target UAV; The channel state information includes all downlink resource elements and corresponds one-to-one with the target reflection path.

2. The UAV perception method for low-altitude environments as described in claim 1, characterized in that, Decoding the 5G downlink signal includes: completing downlink synchronization, orthogonal frequency division multiplexing demodulation, physical channel demodulation, and channel decoding to obtain the downlink data bitstream; and remodulating and mapping the data bitstream according to the 5G physical layer mapping rules to restore the time-frequency resource grid consistent with the transmitter.

3. The UAV perception method for low-altitude environments as described in claim 1, characterized in that, Beamforming is used to form a receiving beam to enhance the reflection path of a target UAV, including: predefining an azimuth angle range, calculating a beamforming weight vector, and forming multiple candidate beams; for each candidate beam, performing element-wise division of the time-frequency resource grid and channel state information reconstruction, calculating the perceived signal-to-noise ratio (SNR) of the target peak in the range-Doppler image, and selecting the candidate beam with the highest perceived SNR as the receiving beam; the perceived SNR is defined as: ,in, and These are the power of the target peak on the range-Doppler plot and the power of the background noise, respectively.

4. The UAV perception method for low-altitude operation as described in claim 3, characterized in that, Calculating the beamforming weight vector includes: beam control based on the antenna array, and pre-setting the angle of arrival. For the distance between For two adjacent antennas, the phase shift of the incident signal is... ;in, It is the carrier wavelength; the steering vector received by the base station, as the beamforming weight vector, is: ,in, It is the total number of antenna elements in the antenna array; by multiplying the received signal at each antenna by the corresponding weight in the steering vector, the incident signal is enhanced and interference is suppressed.

5. A low-altitude UAV perception method as described in claim 1, characterized in that, Based on the range-Doppler image, the distance and radial velocity of the target UAV are detected and estimated, including: detecting target peaks in the range-Doppler image; performing trajectory association using a temporal correlation algorithm on the target peaks detected in multiple consecutive frames to filter out false detections; and outputting estimated UAV distance and velocity values ​​for the associated target trajectory. The UAV distance is: The drone's speed is: ; in, and These are the range cell and Doppler cell indices corresponding to the target peak in the range-Doppler map, respectively; Speed ​​of light; Subcarrier spacing; The carrier wavelength; The duration of the modulation symbol for the s-th carrier; This represents the total number of subcarriers. This represents the total number of carrier modulation symbol frames.

6. The UAV perception method for low-altitude operation as described in claim 1, characterized in that, The channel state information of the target reflection path is reconstructed by performing an element-wise ratio operation between the time-frequency resource cell corresponding to the received beam output signal and the time-frequency resource cell corresponding to the recovered reference transmitted data, including: the reconstructed channel state information satisfies: ,in, It is the time-frequency resource grid that has been restored after decoding. It is the time-frequency resource grid corresponding to the received beam output signal. It is the reconstructed channel state information.

7. A low-altitude UAV perception method as described in claim 1, characterized in that, Generating the range-Doppler map includes: performing a range Fourier transform across subcarriers to convert the frequency-domain channel state information matrix into a range profile, where each path produces a distinct peak at a specific range bin; performing a Doppler Fourier transform across consecutive carrier modulation symbols to convert the phase changes caused by the Doppler frequency shift into a Doppler profile; and combining these two transforms into a joint two-dimensional Fourier transform to obtain the range-Doppler map, as follows: ,in, It is to transform the channel state information matrix to the first... The distance warehouse and the first Fourier transform of a Doppler chamber; This represents the total number of subcarriers. is the total number of carrier modulation symbol frames; k is the subcarrier index, representing the kth subcarrier; s is the carrier modulation symbol frame index, representing the sth carrier modulation symbol frame.

Images