Unmanned aerial vehicle-oriented passive access method and device, electronic equipment and storage medium

Abstract

This application specifically relates to a passive access method, apparatus, electronic device, and storage medium for unmanned aerial vehicles (UAVs). The method includes: determining the data segment to be transmitted in each time slot for a passive terminal; determining the codeword for each time slot based on the data segment to be transmitted in each time slot according to a preset shared codebook tensor, and mapping it to a preset three-dimensional resource tensor to obtain a tensor signal for each time slot; sending the tensor signal for each time slot to a preset center point, and receiving the preprocessed tensor for each time slot sent by the preset center point after tensor reshaping and domain transformation processing; and performing channel tensor recovery processing on the preprocessed tensor for each time slot based on a preset cross-domain iterative detection strategy to obtain the recovered channel for each time slot, and concatenating messages to obtain the recovered data of the passive terminal. This solves the problems of large pilot overhead, complex channel estimation, and transmission instability caused by Doppler shift in high-concurrency UAV scenarios.

Description

Technical Field

[0001] This application relates to the field of signal processing and data transmission technology in communication networks, and in particular to a passive access method, device, electronic device and storage medium for unmanned aerial vehicles (UAVs). Background Technology

[0002] Drones are widely used in logistics, environmental monitoring, emergency rescue, and precision agriculture, and their numbers are experiencing explosive growth. To support the sensing, communication, computing, and control needs of a large number of drones, an efficient and reliable network is required. However, achieving efficient access and accurate positioning for drones faces significant technical challenges.

[0003] Related technologies (such as approximate message passing algorithms) utilize the sparsity of large-scale MIMO channels in the transform domain to reduce detection complexity, while most compressed sensing algorithms in related technologies only process data in the transform-frequency domain.

[0004] However, related technologies (such as approximate message passing algorithms) do not fully utilize multidimensional sparsity. In fact, broadband massive MIMO channels exhibit even stronger sparsity in the transform-delay domain, and channel models in the transform-frequency domain are often difficult to directly embed into classical linear measurement models. This makes it difficult for traditional algorithms to effectively utilize this key feature to improve detection and positioning accuracy, which urgently needs to be addressed. Summary of the Invention

[0005] This application provides a passive access method, device, electronic device, and storage medium for unmanned aerial vehicles (UAVs) to solve problems such as large pilot overhead, complex channel estimation, and unstable transmission caused by Doppler frequency shift in high-concurrency UAV scenarios.

[0006] The first aspect of this application provides a passive access method for unmanned aerial vehicles (UAVs), comprising the following steps: Determine the data segment to be transmitted in each time slot of the passive terminal; Based on a preset shared codebook tensor, the codeword of each time slot is determined according to the data segment to be transmitted in each time slot, and the codeword of each time slot is mapped to a preset three-dimensional resource tensor to obtain the tensor signal of each time slot. Send the tensor signal of each time slot to a preset center point, and receive the preprocessed tensor of each time slot sent by the preset center point after performing tensor reshaping and domain transformation on the tensor signal of each time slot; Based on a preset cross-domain iterative detection strategy, channel tensor recovery processing is performed on the preprocessed tensor of each time slot to obtain the recovery channel of each time slot, and message concatenation is performed according to the recovery channel of each time slot to obtain the recovery data of the passive terminal.

[0007] Optionally, determining the data segment to be transmitted in each time slot sent by the passive terminal includes: Receive data to be transmitted from a passive terminal; Based on a preset time slot partitioning strategy, the data to be transmitted is segmented to obtain the data segment to be transmitted in each time slot.

[0008] Optionally, determining the codeword for each time slot based on a preset shared codebook tensor and the data segment to be transmitted in each time slot includes: The data segment to be transmitted in each time slot is converted into a decimal index to obtain the converted data for each time slot; The codeword sequence corresponding to the transformed data of each time slot is selected from the preset shared codebook tensor to obtain the codeword of each time slot.

[0009] Optionally, the method of performing channel tensor recovery processing on the preprocessed tensor of each time slot based on a preset cross-domain iterative detection strategy to obtain the recovered channel for each time slot includes: The preprocessed tensor of each time slot is initially estimated based on the linear minimum mean square error estimator to obtain a linear estimation result; The linear estimation result is transformed from the transform-frequency domain to the transform-delay domain based on the discrete Fourier transform to obtain the transformed channel tensor; Based on the Bernoulli-Gaussian prior distribution, the transformed channel tensor is estimated element-wise by minimum mean square error to obtain a nonlinear estimation result. Based on the expectation-maximization algorithm, the preset parameters and noise variance in the Bernoulli-Gaussian prior distribution are updated according to the nonlinear estimation results to obtain the updated channel tensor. The updated channel tensor is inversely transformed from the transform-delay domain back to the transform-frequency domain, and the result of the inverse transformation is used as the input for the next linear estimation. The preprocessed tensor of each time slot is pre-estimated based on the linear minimum mean square error estimator, and the estimation result is obtained until the step of inversely transforming the updated channel tensor from the transform-delay domain back to the transform-frequency domain is met, until the preset iteration stop condition is met, and the recovered channel of each time slot is obtained.

[0010] Optionally, the step of concatenating messages according to the recovery channel of each time slot to obtain the recovery data of the passive terminal includes: Detect the energy of each slice in the channel tensor of each recovered channel; Extract the index and slice corresponding to the slice whose energy is greater than a preset threshold, and use the codeword index as the active codeword and the slice as the channel corresponding to the active codeword; Based on the preset k-means clustering algorithm and the channel corresponding to the active codeword, all the active codewords detected in the time slot are clustered to obtain the clustering result; Based on the clustering results, the recovery channels of each time slot arranged in time slot order in the same passive terminal are concatenated to obtain the recovery data of the passive terminal.

[0011] Optionally, the clustering of all active codewords detected in the time slot based on the preset k-means clustering algorithm and the channel corresponding to the active codeword, to obtain the clustering result, includes: The time slot with the most active codewords in all time slots is taken as the reference time slot, the number of active codewords in the reference time slot is taken as the terminal number estimate, and the channel information corresponding to each active codeword in the reference time slot is taken as the current cluster center. For each time slot other than the reference time slot, calculate the Euclidean distance between the channel information amplitude of each active codeword channel in the current time slot and the current cluster center, and construct a cost matrix based on the Euclidean distance; Based on the cost matrix, the Hungarian algorithm is used to assign the active codewords of the current time slot to the corresponding cluster centers to obtain the clustering results.

[0012] A second aspect of this application provides a passive access device for unmanned aerial vehicles (UAVs), comprising: The determination module is used to determine the data segment to be transmitted in each time slot sent by the passive terminal; The mapping module is used to determine the codeword of each time slot based on the data segment to be transmitted in each time slot according to the preset shared codebook tensor, and map the codeword of each time slot to the preset three-dimensional resource tensor to obtain the tensor signal of each time slot. The processing module is used to send the tensor signal of each time slot to a preset center point, and receive the preprocessed tensor of each time slot sent by the preset center point after performing tensor reshaping and domain transformation processing on the tensor signal of each time slot. The recovery module is used to perform channel tensor recovery processing on the preprocessed tensor of each time slot based on a preset cross-domain iterative detection strategy to obtain the recovery channel of each time slot, and to perform message concatenation based on the recovery channel of each time slot to obtain the recovery data of the passive terminal.

[0013] Optionally, the determining module is specifically used for: Receive data to be transmitted from a passive terminal; Based on a preset time slot partitioning strategy, the data to be transmitted is segmented to obtain the data segment to be transmitted in each time slot.

[0014] Optionally, the mapping module is specifically used for: The data segment to be transmitted in each time slot is converted into a decimal index to obtain the converted data for each time slot; The codeword sequence corresponding to the transformed data of each time slot is selected from the preset shared codebook tensor to obtain the codeword of each time slot.

[0015] Optionally, the recovery module is specifically used for: The preprocessed tensor of each time slot is initially estimated based on the linear minimum mean square error estimator to obtain a linear estimation result; The linear estimation result is transformed from the transform-frequency domain to the transform-delay domain based on the discrete Fourier transform to obtain the transformed channel tensor; Based on the Bernoulli-Gaussian prior distribution, the transformed channel tensor is estimated element-wise by minimum mean square error to obtain a nonlinear estimation result. Based on the expectation-maximization algorithm, the preset parameters and noise variance in the Bernoulli-Gaussian prior distribution are updated according to the nonlinear estimation results to obtain the updated channel tensor. The updated channel tensor is inversely transformed from the transform-delay domain back to the transform-frequency domain, and the result of the inverse transformation is used as the input for the next linear estimation. The preprocessed tensor of each time slot is pre-estimated based on the linear minimum mean square error estimator, and the estimation result is obtained until the step of inversely transforming the updated channel tensor from the transform-delay domain back to the transform-frequency domain is met, until the preset iteration stop condition is met, and the recovered channel of each time slot is obtained.

[0016] Optionally, the recovery module is specifically used for: Detect the energy of each slice in the channel tensor of each recovered channel; Extract the index and slice corresponding to the slice whose energy is greater than a preset threshold, and use the codeword index as the active codeword and the slice as the channel corresponding to the active codeword; Based on the preset k-means clustering algorithm and the channel corresponding to the active codeword, all the active codewords detected in the time slot are clustered to obtain the clustering result; Based on the clustering results, the recovery channels of each time slot arranged in time slot order in the same passive terminal are concatenated to obtain the recovery data of the passive terminal.

[0017] Optionally, the recovery module is specifically used for: The time slot with the most active codewords in all time slots is taken as the reference time slot, the number of active codewords in the reference time slot is taken as the terminal number estimate, and the channel information corresponding to each active codeword in the reference time slot is taken as the current cluster center. For each time slot other than the reference time slot, calculate the Euclidean distance between the channel information amplitude of each active codeword channel in the current time slot and the current cluster center, and construct a cost matrix based on the Euclidean distance; Based on the cost matrix, the Hungarian algorithm is used to assign the active codewords of the current time slot to the corresponding cluster centers to obtain the clustering results.

[0018] A third aspect of this application provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being configured to perform a passive access method for unmanned aerial vehicles as described in the above embodiments.

[0019] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the passive access method for unmanned aerial vehicles as described in the above embodiments.

[0020] A fifth aspect of this application provides a computer program product storing a computer program that, when executed by a processor, implements the passive access method for unmanned aerial vehicles as described in the above embodiments.

[0021] Therefore, this embodiment of the application determines the data segment to be transmitted in each time slot of the passive terminal; based on a preset shared codebook tensor, the codeword of each time slot is determined according to the data segment to be transmitted in each time slot, and mapped to a preset three-dimensional resource tensor to obtain the tensor signal of each time slot; the tensor signal of each time slot is sent to a preset center point, and the preset center point performs tensor reshaping and domain transformation processing on the tensor signal of each time slot to send the preprocessed tensor of each time slot; based on a preset cross-domain iterative detection strategy, channel tensor recovery processing is performed on the preprocessed tensor of each time slot to obtain the recovered channel of each time slot, and message concatenation is performed to obtain the recovered data of the passive terminal. This solves the problems of large pilot overhead, complex channel estimation, and transmission instability caused by Doppler shift in high-concurrency UAV scenarios.

[0022] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0023] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a passive access method for unmanned aerial vehicles (UAVs) according to an embodiment of this application; Figure 2 This is a schematic diagram of the time-frequency resource mapping process of a passive access method for unmanned aerial vehicles (UAVs) according to an embodiment of this application; Figure 3 This is a schematic diagram of a cross-domain iterative detection framework for a passive access method for unmanned aerial vehicles (UAVs) according to an embodiment of this application. Figure 4 This is a schematic diagram showing the codeword detection performance of the present invention and a comparative scheme under a common signal-to-noise ratio for a passive access method for unmanned aerial vehicles according to an embodiment of this application. Figure 5 This is a schematic diagram illustrating the message splicing performance of the present invention and a comparative scheme under a common signal-to-noise ratio for a passive access method for unmanned aerial vehicles according to an embodiment of this application. Figure 6 This is a schematic diagram of a passive access device for unmanned aerial vehicles (UAVs) provided according to an embodiment of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application are described in detail below. Examples of these 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. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0025] Before introducing the passive access method for UAVs according to the embodiments of this application, let's briefly introduce the passive access method for UAVs in related technologies.

[0026] Drones are widely used in logistics, environmental monitoring, emergency rescue, and precision agriculture, and their numbers are experiencing explosive growth. To support the sensing, communication, computing, and control needs of a large number of drones, an efficient and reliable network is required. However, achieving efficient access and accurate positioning for drones faces significant technical challenges.

[0027] First, the rapid growth in the number of drones has led to high-concurrency communication. Traditional access methods based on orthogonal resource allocation and "request-authorization" handshake mechanisms not only incur huge signaling overhead but also result in extremely high access latency, making it difficult to meet the real-time requirements of low-altitude services. Therefore, Grant-Free Non-Orthogonal Multiple Access (GF-NOMA) technology has emerged, allowing users to directly transmit data on the same time-frequency resources, becoming the mainstream solution for improving access capacity.

[0028] Secondly, the high mobility of UAVs presents significant communication challenges. The high-speed relative movement between the UAV and the central node generates a significant Doppler shift, causing the wireless channel to exhibit rapidly time-varying characteristics. Current mainstream GF-NOMA technologies typically employ a two-stage active coherent detection scheme: the user first sends its pilot sequence for the central node to estimate the channel, and then sends data symbols for coherent demodulation. However, in highly mobile scenarios, the rapid changes in the channel mean that the channel state information estimated during the pilot stage is often outdated or invalid by the time of data transmission, leading to a sharp decline in data detection performance and compromising transmission reliability. Furthermore, this active access scheme faces complex pilot allocation issues, especially in dynamic situations where UAVs frequently enter and exit the central node's coverage area. The coordination and allocation of pilot resources become extremely complex, easily leading to pilot collisions or resource waste, further limiting the system's flexibility and scalability.

[0029] In addition, in order to achieve passive access identification and location, the receiver usually models joint activity detection, channel estimation and data recovery as a compressed sensing problem to solve.

[0030] While existing techniques (such as approximate message passing algorithms) leverage the sparsity of massive MIMO channels in the transform domain to reduce detection complexity, they do not fully utilize multidimensional sparsity. Most existing compressed sensing algorithms only process data in the transform-frequency domain. In reality, wideband massive MIMO channels exhibit even stronger sparsity in the transform-delay domain. However, channel models in this domain are often difficult to directly embed into classical linear measurement models, making it challenging for traditional algorithms to effectively utilize this crucial feature to improve detection and localization accuracy.

[0031] This application addresses the aforementioned problems by proposing a passive access method for unmanned aerial vehicles (UAVs). In this method, embodiments of this application determine the data segment to be transmitted in each time slot of the passive terminal; based on a preset shared codebook tensor, the codeword for each time slot is determined according to the data segment to be transmitted in each time slot, and mapped to a preset three-dimensional resource tensor to obtain the tensor signal for each time slot; the tensor signal for each time slot is sent to a preset center point, and the preset center point performs tensor reshaping and domain transformation processing on the tensor signal for each time slot, sending the preprocessed tensor for each time slot; based on a preset cross-domain iterative detection strategy, channel tensor recovery processing is performed on the preprocessed tensor for each time slot to obtain the recovered channel for each time slot, and message concatenation is performed to obtain the recovered data of the passive terminal. This solves the problems of large pilot overhead, complex channel estimation, and transmission instability caused by Doppler shift in high-concurrency UAV scenarios.

[0032] Specifically, Figure 1 This is a flowchart illustrating a passive access method for unmanned aerial vehicles (UAVs) provided in an embodiment of this application.

[0033] like Figure 1 As shown, this passive access method for drones includes the following steps: In step S101, the data segment to be transmitted for each time slot sent by the passive terminal is determined.

[0034] Specifically, all drones share a common codebook for non-coherent transmission; in this embodiment, the complete data stream sent by the passive terminal is received, and the data stream is divided into independent data segments corresponding to each time slot according to the time slot structure required for subsequent encoding and transmission, thereby reducing the complexity of single encoding and detection.

[0035] Optionally, in some embodiments, determining the data segment to be transmitted in each time slot sent by the passive terminal includes: receiving the data to be transmitted sent by the passive terminal; and dividing the data to be transmitted into data segments for each time slot based on a preset time slot division strategy.

[0036] Understandably, to mitigate the extremely high detection complexity caused by codebook dimensionality explosion, the terminal (drone) employs a segmented transmission strategy. Assuming each terminal needs to transmit a total of B bits in the current frame, based on a preset time slot division strategy, the transmission frame is divided into T time slots. The terminal then evenly divides the B-bit message sequence into T segments, each containing... Bit data (the data segment to be transmitted in each time slot).

[0037] In step S102, based on a preset shared codebook tensor, the codeword of each time slot is determined according to the data segment to be transmitted in each time slot, and the codeword of each time slot is mapped to a preset three-dimensional resource tensor to obtain the tensor signal of each time slot.

[0038] The preset shared codebook tensor is a set of high-dimensional complex matrices that are pre-generated and stored in all UAV terminals and the central node before communication.

[0039] Specifically, such as Figure 2 As shown, Figure 2 This is a schematic diagram of the time-frequency resource mapping process of a passive access method for unmanned aerial vehicles (UAVs) according to an embodiment of this application; the embodiment of this application presupposes a shared codebook tensor known throughout the network. ,in The codebook size is [size]. The UAV maps selected codewords to a physical layer 3D resource tensor, obtaining the tensor signal for each time slot. This resource tensor consists of G resource blocks. Each resource block contains [data / information]. consecutive subcarriers and time domain OFDM symbols, and satisfying The mapped signal is modulated by OFDM (Orthogonal Frequency Division Multiplexing) and then transmitted to the central node through a time-frequency dual-selection channel.

[0040] Optionally, in some embodiments, the codeword for each time slot is determined based on a preset shared codebook tensor according to the data segment to be transmitted in each time slot, including: converting the data segment to be transmitted in each time slot into a decimal index to obtain the converted data of each time slot; and selecting the codeword sequence corresponding to the converted data of each time slot from the preset shared codebook tensor to obtain the codeword for each time slot.

[0041] Understandably, in the t-th time slot, the terminal converts the current J bits of data into a decimal index and selects the corresponding codeword sequence from the shared codebook. This approach avoids the waste of resources that would otherwise be allocated to a massive number of dedicated codebooks for a large number of users.

[0042] In step S103, the tensor signal of each time slot is sent to a preset center point, and the preprocessed tensor of each time slot is received from the preset center point after performing tensor reshaping and domain transformation processing on the tensor signal of each time slot.

[0043] Specifically, the tensor signal received by the central node is the superposition of signals transmitted by all active terminals. The central node utilizes the approximately flat fading characteristic of the channel within the same resource block to reshape the received signal tensor; assuming the channel within the g-th resource block remains unchanged, the received signal is reshaped as follows: This adds a simplified system model that facilitates codeword detection by the central node. The central node leverages the sparsity of the broadband massive MIMO channel in the transform-delay domain, using a discrete Fourier transform matrix to transform the reshaped received signal from the spatial-frequency domain to the transform-frequency domain (transform domain). This aims to reduce the density of non-zero elements in the channel tensor to be estimated, thereby improving the recovery performance of subsequent compressed sensing algorithms.

[0044] In step S104, based on the preset cross-domain iterative detection strategy, the preprocessed tensor of each time slot is processed by channel tensor recovery to obtain the recovery channel of each time slot, and the recovery data of the passive terminal is obtained by message splicing according to the recovery channel of each time slot.

[0045] Specifically, such as Figure 3 As shown, Figure 3This diagram illustrates a cross-domain iterative detection framework for a passive access method for unmanned aerial vehicles (UAVs) according to an embodiment of this application. The central node uses the Cross-Domain Orthogonal Approximate Message Passing for Generalized Multiple Measurement Vector (CD-OAMP-GMMV) algorithm (a preset cross-domain iterative detection strategy) to recover the channel tensor. Within the OAMP framework, this algorithm iterates alternately between the transform-frequency domain and the transform-delay domain, and incorporates the Expectation Maximization (EM) algorithm to adaptively learn channel parameters, fully utilizing the structured sparsity of wideband massive MIMO channels in the angle-delay domain. After recovering the channel tensor, the central node needs to reconstruct the complete message belonging to the same UAV (recovered data from the passive terminal) using message splicing technology.

[0046] Optionally, in some embodiments, based on a preset cross-domain iterative detection strategy, channel tensor recovery processing is performed on the preprocessed tensor of each time slot to obtain the recovered channel for each time slot, including: performing a preliminary estimation of the preprocessed tensor of each time slot based on a linear minimum mean square error estimator to obtain a linear estimation result; transforming the linear estimation result from the transform-frequency domain to the transform-delay domain based on the discrete Fourier transform to obtain the transformed channel tensor; performing element-wise minimum mean square error estimation on the transformed channel tensor based on the Bernoulli-Gaussian prior distribution to obtain a nonlinear estimation result; and according to the expected... The maximization algorithm updates the preset parameters and noise variance in the Bernoulli-Gaussian prior distribution based on the nonlinear estimation results, obtaining the updated channel tensor. The updated channel tensor is then inversely transformed from the transform-delay domain back to the transform-frequency domain, and the result of the inverse transformation is used as the input for the next linear estimation. The process of performing a preliminary estimation of the preprocessed tensor for each time slot based on the linear minimum mean square error estimator is repeated until the updated channel tensor is inversely transformed from the transform-delay domain back to the transform-frequency domain, until the preset iteration stopping condition is met, thus obtaining the recovered channel for each time slot.

[0047] It is understandable that the estimated values ​​of the initial channel tensor and the noise variance are... And prior parameters (such as mean, variance, and sparsity). In the transform-frequency domain, a preliminary estimate of the channel tensor is performed using a linear minimum mean square error (LMMSE) estimator to obtain a linear estimation result; the central node calculates the linear estimation result and its error variance, and uses a discrete Fourier transform to transform the output of the linear estimation from the transform-frequency domain to the transform-delay domain. In the angle-delay domain, the channel exhibits the strongest sparsity.

[0048] Specifically, this embodiment of the application, based on the Bernoulli-Gaussian prior distribution, performs element-wise minimum mean square error (MMSE) estimation on the transformed channel tensor and calculates the posterior mean. The Expectation Maximization (EM) algorithm is used to update the parameters (sparse probability, mean, variance) and noise in the Bernoulli-Gaussian prior distribution using the estimated values ​​from the current iteration. In particular, the nearest neighbor sparse pattern learning method is used to update the sparse probability parameters. The prior sparse probability of each element is updated by calculating the average sparse probability of each element in its neighborhood in the transform domain (adjacent transform grid points) and the delay domain (adjacent sampling points). This embodiment of the application uses the Discrete Fourier Transform to inversely transform the denoised channel tensor obtained from the nonlinear estimation from the transform-delay domain back to the transform-frequency domain. Simultaneously, the error variance of the nonlinear estimation in the angle-frequency domain is calculated. The output (the result of the inverse transformation) is then used as the input for the next iteration of linear estimation. In addition, to prevent the algorithm from diverging, this embodiment of the application introduces a damping factor to smoothly update the input of the current iteration.

[0049] Furthermore, the initial estimation is repeated until the updated channel tensor is inversely transformed from the transform-delay domain back to the transform-frequency domain, until the maximum number of iterations is reached or the error between two adjacent iterations is less than a preset error threshold. Finally, the converged channel tensor estimate is output.

[0050] Optionally, in some embodiments, message concatenation is performed based on the recovery channels of each time slot to obtain the recovery data of the passive terminal, including: detecting the energy of each slice in the channel tensor of each recovery channel; extracting the index and slice corresponding to the slice whose energy is greater than a preset threshold, using the codeword index as the active codeword, and the slice as the channel corresponding to the active codeword; clustering all active codewords detected in the time slot based on a preset k-means clustering algorithm and the channel corresponding to the active codeword to obtain the clustering result; and concatenating the recovery channels of each time slot arranged in time slot order in the same passive terminal based on the clustering result to obtain the recovery data of the passive terminal.

[0051] The preset threshold can be a threshold set by the user, a threshold obtained through a limited number of experiments, or a threshold obtained through a limited number of computer simulations; no specific limitation is made here.

[0052] Understandably, the central node detects and calculates the energy of each slice of the transformed-delay domain channel tensor recovered for each time slot; it extracts the references of slices with energy greater than a certain threshold, as well as the slices themselves, and uses them as active codewords and their corresponding channels, respectively. Since the transformed domain and delay information of the channel remain relatively stable across different time slots within the same frame for the same UAV, this invention utilizes this characteristic to propose an improved k-means clustering algorithm (a pre-defined k-means clustering algorithm) for message concatenation. This algorithm concatenates the recovered channels of each time slot arranged in time slot order within the same passive terminal to obtain the recovered data of the passive terminal.

[0053] Optionally, in some embodiments, based on a preset k-means clustering algorithm and the channels corresponding to active codewords, all active codewords detected in the time slots are clustered to obtain clustering results. This includes: using the time slot with the most active codewords among all time slots as a reference time slot, using the number of active codewords in the reference time slot as an estimate of the number of terminals, and using the channel information corresponding to each active codeword in the reference time slot as the current cluster center; for each time slot other than the reference time slot, calculating the Euclidean distance between the channel of each active codeword in the current time slot and the channel information amplitude of the current cluster center, and constructing a cost matrix based on the Euclidean distance; and using the Hungarian algorithm to assign the active codewords of the current time slot to the corresponding cluster centers according to the cost matrix to obtain clustering results.

[0054] Understandably, among all time slots, the time slot with the most active codewords is selected as the reference time slot. The number of codewords is used as an estimate of the number of active UAVs (terminal count estimate), and the channel corresponding to each codeword is used as the initial cluster center. Subsequently, codewords and channels are processed one time slot at a time; a cost matrix is ​​constructed, whose elements are the Euclidean distance between the channel amplitude of each active codeword in a given time slot and the channel amplitude of the current cluster center. Based on the cost matrix, the Hungarian algorithm is used to assign channels in each time slot to their corresponding clusters. Codewords corresponding to channels in each cluster are considered to belong to the same UAV. If the number of codewords in the currently processed time slot is less than the number of cluster centers, the codeword with the largest minimum distance to each cluster center is selected and identified as a duplicate codeword. The rows of the cost matrix corresponding to this codeword are copied and concatenated with the original cost matrix, and then the concatenated cost matrix is ​​processed using the Hungarian algorithm. The cost matrix calculation and clustering and concatenation steps are executed iteratively until the clustering results are the same in two adjacent iterations, thus obtaining the clustering result.

[0055] Finally, the codeword indices arranged in chronological order in each cluster are extracted and concatenated into a complete bit data stream, thereby completing the data recovery of the passive terminal.

[0056] It should be noted that by utilizing the high-precision channel state information recovered in the angle-delay domain, the central node can directly extract the direction of arrival (DOA) and delay parameters of the active terminal, thereby achieving direction finding and positioning of the UAV.

[0057] To facilitate a better understanding of the passive access method for unmanned aerial vehicles (UAVs) according to the embodiments of this application by those skilled in the art, the following is combined with... Figures 2 to 5 The embodiments shown will be described in detail.

[0058] Specifically, this invention discloses a passive access identification and direction finding method for unmanned aerial vehicles (UAVs), aiming to meet the identification and data transmission needs of large-scale terminals under the coverage of a central node. It achieves efficient detection by leveraging the sparsity of the channel in the transform-delay domain through the collaborative mapping of time and frequency domain resources. This embodiment considers a scenario where a central node equipped with multiple ports serves multiple UAVs. The number of ports in the central node is... A uniform planar array; the total number of UAVs is K, and OFDM modulation is used for uplink transmission to counteract frequency-selective fading of the channel, with a total number of OFDM subcarriers of m.

[0059] This embodiment considers a time-frequency dual-selection channel model and uses... The channel vector of UAV with sequence number k on the m-th subcarrier and the l-th OFDM symbol is expressed as: ; The specific expression for the phase term is as follows: ; in, and These are the azimuth and pitch angles of arrival for the UAV with serial number k to the central node; each terminal's channel contains P paths; , and These are the complex gain, delay, and Doppler shift of the p-th path corresponding to the terminal with serial number k; , It is the large-scale fading factor and time synchronization error of the terminal with serial number k; and These represent the system bandwidth and the OFDM (Orthogonal Frequency Division Multiplexing) subcarrier spacing, respectively. This is the guide vector for the central node. This guide vector can be represented as: ; in , ; , , For wavelength, It refers to the port spacing.

[0060] Furthermore, embodiments of this application include the following steps: Step 1: Message Segmentation and Encoding Transmission To mitigate the extremely high detection complexity caused by codebook dimensionality explosion, the UAV employs a segmented transmission strategy. Assume that each UAV needs to transmit a total of B bits in the current frame. The system divides this transmission frame into T time slots. The UAV evenly divides the B-bit message sequence into T segments, each segment containing... Bit data. The system pre-defines a shared codebook tensor known across the entire network. ,in Let J be the codebook size. In the t-th time slot, the terminal converts the current J bits of data into a decimal index and selects the corresponding codeword sequence from the shared codebook. To reduce the complexity of the detection method, the vectors in the codebook of this invention... All are generated by randomly selecting row vectors from the discrete Fourier transform matrix.

[0061] Step 1.2: The central node pre-allocates a set of contiguous OFDM resource blocks, which are constructed as a three-dimensional resource tensor at the physical layer, with the dimension defined as follows: Its form is shown in Figure 1. Where G is the total number of resource blocks, The number of consecutive subcarriers contained in each resource block. The number of consecutive OFDM symbols occupied for each resource block; and satisfy The interval between adjacent resource blocks is... There are subcarriers. Let the relative time-frequency offset on the g-th resource block in the t-th time slot be denoted as . The channel vector on the subcarrier is .

[0062] Furthermore, the drone will use the dimension selected in step 1.1 as... The code words (of which) The time-frequency signal is mapped to the resource block mentioned above. After being modulated by OFDM, the mapped time-frequency signal is transmitted through the N ports of the terminal.

[0063] Furthermore, the signal received by the central node is the result of the superposition of signals transmitted by all active terminals. After passing through the time-frequency dual-selection channel, the central node receives the signal in the g-th subcarrier group. The signal received on each subcarrier It can be represented as: ; Among them, tensor This represents the codebook after time-frequency resource mapping. (Vector) This is a one-hot or zero vector used to represent the codeword selection operation of the UAV. When the UAV with sequence number k does not transmit data, It is a zero vector; when the drone with sequence number k transmits data, It is a one-hot vector that stores the sequence number of the codeword it transmits. This represents Gaussian white noise.

[0064] Step 2: Tensor reshaping and domain transformation preprocessing of the received signal at the central node.

[0065] Step 2.1 Tensor Reshaping: The central node reshapes the received signal into a tensor to perform codeword detection using a simplified system model.

[0066] Directly performing codeword detection based on the system model with a time-frequency dual-selection channel as described in the above formula is quite difficult. However, the time-frequency resource mapping scheme proposed in this chapter can flexibly adjust the time resources and frequency bandwidth occupied by the transmitted codewords. After ensuring that the time span of each resource block is less than the channel's coherence time and its occupied bandwidth is less than the channel's coherence bandwidth, we can reasonably assume that the channel in each... The resource block exhibits locally flat fading characteristics. Therefore, the system model for codeword detection on the g-th resource block can be expressed as the following formula: ; in, Let be the locally flat fading channel of the k-th UAV on the g-th resource block, as assumed. Rearranging the above formulas into matrix form yields a simplified system model for codeword detection: ; in, Therefore, the matrix The row vector is non-zero only when its row number corresponds to the codeword selected by the user; otherwise, it is a zero vector. This makes... It exhibits sparsity. Therefore, the tensor of the central node can be recovered using compressed sensing-like algorithms. Data detection is achieved by detecting the positions of non-zero rows. It can be obtained by tensor reshaping operations, which are as follows: .

[0067] Step 2.2, Domain Transformation: The central node utilizes the sparsity of the broadband massive MIMO channel in the transform-delay domain, using a discrete Fourier transform matrix to transform the reshaped received signal from the spatial-frequency domain to the transform-frequency domain (transform domain). This step aims to reduce the density of non-zero elements in the channel tensor to be estimated, thereby improving the recovery performance of subsequent compressed sensing algorithms.

[0068] Step 3: Channel Tensor Recovery Based on Cross-Domain Iterative Detection (CD-OAMP-GMMV) The central node employs the Cross-Domain Orthogonal Approximate Message Passing for Generalized Multiple Measurement Vector (CD-OAMP-GMMV) algorithm to recover the channel tensor. Within the OAMP framework, this algorithm iterates alternately between the transform-frequency domain and the transform-delay domain, and incorporates the Expectation Maximization (EM) algorithm to adaptively learn channel parameters, fully utilizing the structured sparsity of wideband massive MIMO channels in the angle-delay domain. The specific process is as follows: Step 3.1, Initialization: Initialize the estimated values ​​of the channel tensor and the noise variance. And prior parameters (such as mean, variance, and sparsity).

[0069] Step 3.2, Linear Estimation: In the transform-frequency domain, a preliminary estimate of the channel tensor is performed using a Linear Minimum Mean Square Error (LMMSE) estimator. This step aims to decouple the high-dimensional matrix estimation problem and eliminate the correlation of the observation matrices. The central node calculates the output of the linear estimate. and its error variance Due to the partial orthogonality of the shared codebook matrix, complex matrix inversion operations are avoided, reducing computational complexity. The calculation formula is as follows:

[0070] ; Step 3.3, Domain Transformation: The output of the linear estimate is transformed from the transform-frequency domain to the transform-delay domain using the Discrete Fourier Transform module. Due to the orthogonalization employed, the output of the linear estimate can be considered as an independent Gaussian distribution. Utilizing the linearity of independent Gaussian distributions, the following transformation process can be obtained:

[0071] ; in, and It is the result and error of the linear estimation transformed to the transform-delay domain. It is a DFT matrix. Represents a diagonal matrix. This represents the element in the m-th row and n-th column of the matrix.

[0072] Step 3.4, Non-Linear Estimator (NLE) and Parameter Learning: In the angle-delay domain, the channel exhibits the strongest sparsity. This step is processed through the following three sub-modules: MMSE denoising: Based on the Bernoulli-Gaussian prior distribution, the transformed channel tensor... Perform element-wise MMSE estimation. The prior distribution is: ; The posterior distribution is: ; The specific expressions for each parameter are as follows: , , ; The channel estimation results and errors in the transform-delay domain are as follows:

[0073] .

[0074] Finally, the channel estimation results are orthogonalized to decouple the estimation error from the signal. The nonlinear estimation output is then fed into the transform-delay domain estimation results for the next iteration. for: ; in, These are the orthogonalization parameters.

[0075] Step 3.4, Parameter Adaptive Learning: In the actual system, the channel prior parameters (such as the mean in the prior distribution) ,variance sparsity ) and system noise variance These parameters are often difficult to obtain. Therefore, this invention employs the Expectation Maximization (EM) algorithm to learn these parameters in order to improve estimation accuracy.

[0076] , , .

[0077] To leverage the clustered sparsity of the channel in the angle-delay domain, this invention introduces a nearest-neighbor sparse pattern learning mechanism. The prior sparsity probability of each element is updated by calculating its average sparsity in the transform domain (adjacent grid points) and the delay domain (adjacent sampling points) neighborhood. ; in, coordinates The neighbor set contains six elements, specifically, two adjacent grid points in the delay domain and two adjacent grid points in the angular domain along the x-axis and y-axis of the uniform planar array.

[0078] Step 3.5, Inverse Domain Transformation: The denoised channel tensor obtained from the nonlinear estimation is inversely transformed from the angle-delay domain back to the angle-frequency domain, denoted as... .

[0079] ; Furthermore, the error of the nonlinear estimation in the angle-frequency domain is calculated: .

[0080] Step 3.6, Iteration and Damped Update: The output of Step 3.5 is used as the input for the next iterative linear estimation. To prevent the algorithm from diverging, a damping factor is introduced. The input for this iteration is then updated smoothly.

[0081] Repeat steps 3.2 to 3.6 until the maximum number of iterations I is reached or the normalized mean square error between two adjacent iterations is less than a preset threshold. The final output is the converged transform-delay domain channel tensor estimate.

[0082] Step 4: Active codeword detection and message concatenation.

[0083] After recovering the channel tensor, the central node first detects the codewords sent by the UAV in each time slot, and then needs to use message splicing technology to restore the complete message belonging to the same UAV.

[0084] Step 4.1, Central Node Detects Active Terminals: Further, calculate the square of the Frobenius norm of each slice corresponding to each codeword in the channel tensor, i.e.: ; If this value is greater than a certain threshold, it is assumed that the corresponding codeword was sent by a drone.

[0085] Step 4.2 Message concatenation based on channel feature clustering: Because the channel transform domain and delay information of the same UAV remain relatively stable across multiple time slots within the same frame, this invention utilizes this characteristic to propose an improved time-slot-wise k-means clustering algorithm for message concatenation. Within each time slot, the codeword concatenation problem can be modeled as an optimal allocation problem, with the optimization objective being to minimize the sum of distances from the codewords in each time slot to each cluster center, which can be solved using the Hungarian algorithm. The message concatenation process based on the k-means algorithm proposed in this invention is as follows: Cluster center initialization: Among all time slots, select the time slot with the most active codewords, and use the number of codewords in that time slot as an estimate of the number of active drones. ; in Let be the estimated number of codewords for the t-th time slot. Then, use the channels corresponding to the active codewords in this time slot as the initial cluster centers.

[0086] ; Subsequently, codewords and channels are processed one time slot at a time.

[0087] Cost matrix calculation: The cost matrix is ​​the input to the Hungarian algorithm and contains the distances from the codewords in the current time slot to each cluster center. In the cost matrix constructed in this invention, each element is the Euclidean distance between the channel amplitude of an active codeword in time slot t and the channel amplitude of the current cluster center.

[0088] ; Using amplitude characteristics can effectively eliminate the rapid rotation effect on the phase caused by the Doppler frequency shift due to drone movement.

[0089] Clustering and concatenation: Using either the Hungarian algorithm or a greedy algorithm, the active codewords detected in this time slot are divided into cluster centers, minimizing the sum of the distances from each codeword to its corresponding cluster center. The output of the Hungarian algorithm is a binary matrix. ,in This indicates whether the i-th codeword belongs to the k-th cluster center. Based on the clustering results, the codewords and channels included in each category are updated.

[0090] .

[0091] Cluster center update: The mean of the channel slices corresponding to the codewords in each category is used as the new cluster center.

[0092] The cost matrix calculation and clustering and splicing steps are performed repeatedly until the clustering results are the same in two adjacent loops.

[0093] Data recovery: Extract the codeword indexes arranged in chronological order from each cluster, convert them into binary, and concatenate them into a complete bit data stream to complete the data recovery and identity verification of the drone (if the data contains an identity ID segment).

[0094] Step 4.3, Direction Finding and Positioning: Using the high-precision channel state information recovered in the transform-delay domain in Step 3, the central node can extract the direction of arrival and delay parameters of the active terminal by searching the index of the channel peak in the transform-delay domain, thereby realizing the direction finding and positioning of the UAV.

[0095] Therefore, the embodiments of this application solve the problems of excessive pilot overhead and high channel estimation complexity caused by coherent detection in UAV communication, as well as the stability of data transmission due to Doppler shift interference in scenarios with rapidly changing channels. Unlike traditional pilot-dependent transmission modes, this invention innovatively proposes a codebook-based noncoherent transmission strategy. First, by constructing a high-dimensional tensor codebook known to both the terminal and the central node, the data bits to be transmitted are directly mapped to sequence indices in the codebook, thereby achieving joint encoding of data information and terminal identity, enabling the receiver to break free from strong dependence on instantaneous channel state information. Based on this, this invention designs a time-frequency resource mapping mechanism, mapping the encoded sequence to multi-dimensional time-frequency resource blocks, reducing the time resources occupied by the signal to cope with the impact of time-varying channels. Furthermore, the cross-domain iterative detection method introduced at the receiver in this invention enables the invention to fully utilize the sparse characteristics of the channel in the transform-delay domain without explicit channel estimation, achieving highly reliable, low-latency concurrent access for massive numbers of terminals in complex, highly mobile environments.

[0096] Furthermore, the embodiments of this application can also verify the results of the passive access method for unmanned aerial vehicles.

[0097] The number of active drones is 20, and the number of central node ports is 8×8=64. Codeword length Number of subcarrier groups The following two comparison algorithms were selected to verify the advantages of this invention in UAV communication scenarios.

[0098] (1) A passive access scheme based on the Simultaneous Orthogonal Matching Pursuit (SOMP) algorithm; channel tensor recovery is performed using the SOMP algorithm. The present invention is compared with this scheme to demonstrate the advantages of the cross-domain iterative detection mechanism proposed in this invention in terms of data detection performance.

[0099] (2) A passive access scheme based on a multi-vector observation approximate message passing algorithm; using the GMMV-AMP algorithm to recover the channel tensor at the transform-frequency. The present invention is compared with this scheme to demonstrate the advantages of the cross-domain iterative detection mechanism proposed in this invention in terms of data detection performance.

[0100] The accuracy of codeword detection and message concatenation is measured using Code Detection Error Probability (CDEP) and Per-User Probability of Error (PUPE), respectively. CDEP and PUPE are calculated as follows:

[0101] ; in, and Let represent the number of codewords missed and falsely detected in the t-th time slot, respectively. and These represent the number of messages missed and the number of messages that were falsely detected after message concatenation, respectively.

[0102] Furthermore, such as Figure 4 and Figure 5 As shown, Figure 4 This is a schematic diagram illustrating the codeword detection performance of the present invention and a comparative scheme under a common signal-to-noise ratio for a passive access method for unmanned aerial vehicles according to an embodiment of this application. Figure 5 This diagram illustrates the message concatenation performance of the proposed method and a comparative scheme under common signal-to-noise ratios (SNRs) for a passive access method for unmanned aerial vehicles (UAVs) according to an embodiment of this application. The CDEP and PUPE of different schemes under common SNRs are compared. It can be seen that the proposed scheme outperforms the other two comparative schemes in both CDEP and PUPE. This is because the proposed scheme employs a cross-domain iterative detection mechanism, which can utilize the stronger sparsity of the UAV channel in the transform-delay domain, thereby enhancing the accuracy of channel tensor recovery. Furthermore, the more accurate channel tensor recovery makes codeword detection and concatenation more accurate. In contrast, the comparative schemes can only utilize the sparsity of the channel in the transform domain, resulting in poor channel tensor recovery performance.

[0103] Therefore, addressing the issues of fast time-varying channel characteristics caused by the high mobility of UAVs and the huge pilot overhead in traditional active coherent detection, this invention first utilizes a shared codebook tensor known throughout the network on the UAV side to directly map the segmented data information into sequence indices, which are then mapped to multi-dimensional time-frequency resource blocks for transmission. At the receiving end, the central node reshapes the received signal into a tensor and, leveraging the structured sparsity of the channel in the transform-delay domain, employs a cross-domain orthogonal approximate message passing-generalized multi-measurement vector algorithm to recover the channel tensor, while simultaneously combining an expectation-maximization algorithm to adaptively learn channel parameters. After channel recovery, an improved channel feature-based clustering algorithm is used to complete message splicing and identity recognition, and the direction of arrival and delay parameters are directly extracted from the recovered channel state information to achieve accurate direction finding and positioning of the UAV. This invention can effectively improve the reliability and detection accuracy of concurrent access for massive numbers of UAV terminals in high-mobility and high-concurrency scenarios.

[0104] According to the passive access method for UAVs proposed in this application, the method determines the data segment to be transmitted in each time slot of the passive terminal; based on a preset shared codebook tensor, the codeword of each time slot is determined according to the data segment to be transmitted in each time slot, and mapped to a preset three-dimensional resource tensor to obtain the tensor signal of each time slot; the tensor signal of each time slot is sent to a preset center point, and the preprocessed tensor of each time slot is received from the preset center point after tensor reshaping and domain transformation processing of the tensor signal of each time slot; based on a preset cross-domain iterative detection strategy, channel tensor recovery processing is performed on the preprocessed tensor of each time slot to obtain the recovered channel of each time slot, and message concatenation is performed to obtain the recovered data of the passive terminal. This solves the problems of large pilot overhead, complex channel estimation, and transmission instability caused by Doppler shift in high-concurrency UAV scenarios.

[0105] Next, referring to the accompanying drawings, a passive access device for unmanned aerial vehicles (UAVs) according to an embodiment of this application is described.

[0106] Figure 6 This is a block diagram of a passive access device for unmanned aerial vehicles according to an embodiment of this application.

[0107] like Figure 6 As shown, the passive access device 10 for UAVs includes: a determination module 100, a mapping module 200, a processing module 300, and a recovery module 400.

[0108] The determining module 100 is used to determine the data segment to be transmitted in each time slot sent by the passive terminal. The mapping module 200 is used to determine the codeword of each time slot based on the data segment to be transmitted in each time slot according to the preset shared codebook tensor, and map the codeword of each time slot to the preset three-dimensional resource tensor to obtain the tensor signal of each time slot. The processing module 300 is used to send the tensor signal of each time slot to a preset center point, and receive the preprocessed tensor of each time slot sent by the preset center point after performing tensor reshaping and domain transformation processing on the tensor signal of each time slot. The recovery module 400 is used to perform channel tensor recovery processing on the preprocessed tensor of each time slot based on a preset cross-domain iterative detection strategy to obtain the recovery channel of each time slot, and to perform message concatenation based on the recovery channel of each time slot to obtain the recovery data of the passive terminal.

[0109] Optionally, the determining module 100 is specifically used for: receiving data to be transmitted sent by a passive terminal; and dividing the data to be transmitted into data segments for each time slot based on a preset time slot division strategy.

[0110] Optionally, the mapping module 200 is specifically used to: convert the data segment to be transmitted in each time slot into a decimal index to obtain the converted data of each time slot; and select the codeword sequence corresponding to the converted data of each time slot from a preset shared codebook tensor to obtain the codeword of each time slot.

[0111] Optionally, the recovery module 400 is specifically used for: performing a preliminary estimation of the preprocessed tensor of each time slot based on a linear minimum mean square error estimator to obtain a linear estimation result; transforming the linear estimation result from the transform-frequency domain to the transform-delay domain based on the discrete Fourier transform to obtain the transformed channel tensor; performing element-wise minimum mean square error estimation on the transformed channel tensor based on the Bernoulli-Gaussian prior distribution to obtain a nonlinear estimation result; updating the preset parameters and noise variance in the Bernoulli-Gaussian prior distribution based on the expectation-maximization algorithm and the nonlinear estimation result to obtain the updated channel tensor; inversely transforming the updated channel tensor from the transform-delay domain back to the transform-frequency domain, and using the result of the inverse transformation as the input for the next linear estimation, repeating the steps of performing a preliminary estimation of the preprocessed tensor of each time slot based on the linear minimum mean square error estimator to obtain the estimation result until the updated channel tensor is inversely transformed from the transform-delay domain back to the transform-frequency domain, until the preset iteration stopping condition is met, to obtain the recovered channel for each time slot.

[0112] Optionally, the recovery module 400 is specifically used for: detecting the energy of each slice in the channel tensor of each recovery channel; extracting the index and slice corresponding to the slice whose energy is greater than a preset threshold, using the codeword index as the active codeword, and the slice as the channel corresponding to the active codeword; clustering all active codewords detected in the time slot based on the preset k-means clustering algorithm and the channel corresponding to the active codeword, and obtaining the clustering result; and concatenating the recovery channels of each time slot arranged in time slot order in the same passive terminal with messages based on the clustering result, to obtain the recovery data of the passive terminal.

[0113] Optionally, the recovery module 400 is specifically used to: take the time slot with the most active codewords among all time slots as the reference time slot, use the number of active codewords in the reference time slot as the terminal quantity estimate, and use the channel information corresponding to each active codeword in the reference time slot as the current cluster center; for each time slot other than the reference time slot, calculate the Euclidean distance between the channel information amplitude of each active codeword in the current time slot and the channel information amplitude of the current cluster center, and construct a cost matrix based on the Euclidean distance; according to the cost matrix, use the Hungarian algorithm to assign the active codewords of the current time slot to the corresponding cluster centers to obtain the clustering results.

[0114] It should be noted that the foregoing explanation of the passive access method embodiment for UAVs also applies to the passive access device for UAVs in this embodiment, and will not be repeated here.

[0115] According to the passive access device for UAVs proposed in this application, the embodiments of this application determine the data segment to be transmitted in each time slot of the passive terminal; based on a preset shared codebook tensor, the codeword of each time slot is determined according to the data segment to be transmitted in each time slot, and mapped to a preset three-dimensional resource tensor to obtain the tensor signal of each time slot; the tensor signal of each time slot is sent to a preset center point, and the preset center point performs tensor reshaping and domain transformation processing on the tensor signal of each time slot to send the preprocessed tensor of each time slot; based on a preset cross-domain iterative detection strategy, the preprocessed tensor of each time slot is processed by channel tensor recovery to obtain the recovered channel of each time slot, and message concatenation is performed to obtain the recovered data of the passive terminal. Thus, the problems of large pilot overhead, complex channel estimation, and transmission instability caused by Doppler shift are solved in high-concurrency UAV scenarios.

[0116] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0117] When the processor 702 executes the program, it implements the passive access method for UAVs provided in the above embodiments.

[0118] Furthermore, electronic devices also include: Communication interface 703 is used for communication between memory 701 and processor 702.

[0119] The memory 701 is used to store computer programs that can run on the processor 702.

[0120] The memory 701 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0121] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0122] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0123] The processor 702 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0124] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described passive access method for unmanned aerial vehicles.

[0125] This application also provides a computer program product that stores a computer program that, when executed by a processor, implements the above-described passive access method for unmanned aerial vehicles.

[0126] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0127] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0128] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0129] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0130] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

Claims

1. A passive access method for unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: Determine the data segment to be transmitted in each time slot of the passive terminal; Based on a preset shared codebook tensor, the codeword of each time slot is determined according to the data segment to be transmitted in each time slot, and the codeword of each time slot is mapped to a preset three-dimensional resource tensor to obtain the tensor signal of each time slot. Send the tensor signal of each time slot to a preset center point, and receive the preprocessed tensor of each time slot sent by the preset center point after performing tensor reshaping and domain transformation on the tensor signal of each time slot; Based on a preset cross-domain iterative detection strategy, channel tensor recovery processing is performed on the preprocessed tensor of each time slot to obtain the recovery channel of each time slot, and message concatenation is performed according to the recovery channel of each time slot to obtain the recovery data of the passive terminal.

2. The method according to claim 1, characterized in that, The step of determining the data segment to be transmitted in each time slot sent by the passive terminal includes: Receive data to be transmitted from a passive terminal; Based on a preset time slot partitioning strategy, the data to be transmitted is segmented to obtain the data segment to be transmitted in each time slot.

3. The method according to claim 1 or 2, characterized in that, The method of determining the codeword for each time slot based on the preset shared codebook tensor and the data segment to be transmitted in each time slot includes: The data segment to be transmitted in each time slot is converted into a decimal index to obtain the converted data for each time slot; The codeword sequence corresponding to the transformed data of each time slot is selected from the preset shared codebook tensor to obtain the codeword of each time slot.

4. The method according to claim 1, characterized in that, The method based on a preset cross-domain iterative detection strategy performs channel tensor recovery processing on the preprocessed tensor of each time slot to obtain the recovered channel for each time slot, including: The preprocessed tensor of each time slot is initially estimated based on the linear minimum mean square error estimator to obtain a linear estimation result; The linear estimation result is transformed from the transform-frequency domain to the transform-delay domain based on the discrete Fourier transform to obtain the transformed channel tensor; Based on the Bernoulli-Gaussian prior distribution, the transformed channel tensor is estimated element-wise by minimum mean square error to obtain a nonlinear estimation result. Based on the expectation-maximization algorithm, the preset parameters and noise variance in the Bernoulli-Gaussian prior distribution are updated according to the nonlinear estimation results to obtain the updated channel tensor. The updated channel tensor is inversely transformed from the transform-delay domain back to the transform-frequency domain, and the result of the inverse transformation is used as the input for the next linear estimation. The preprocessed tensor of each time slot is pre-estimated based on the linear minimum mean square error estimator, and the estimation result is obtained until the step of inversely transforming the updated channel tensor from the transform-delay domain back to the transform-frequency domain is met, until the preset iteration stop condition is met, and the recovered channel of each time slot is obtained.

5. The method according to claim 1, characterized in that, The step of concatenating messages according to the recovery channel of each time slot to obtain the recovery data of the passive terminal includes: Detect the energy of each slice in the channel tensor of each recovered channel; Extract the index and slice corresponding to the slice whose energy is greater than a preset threshold, and use the codeword index as the active codeword and the slice as the channel corresponding to the active codeword; Based on the preset k-means clustering algorithm and the channel corresponding to the active codeword, all the active codewords detected in the time slot are clustered to obtain the clustering result; Based on the clustering results, the recovery channels of each time slot arranged in time slot order in the same passive terminal are concatenated to obtain the recovery data of the passive terminal.

6. The method according to claim 5, characterized in that, The method, based on a preset k-means clustering algorithm and the channel corresponding to the active codeword, clusters all the active codewords detected in the time slot to obtain clustering results, including: The time slot with the most active codewords in all time slots is taken as the reference time slot, the number of active codewords in the reference time slot is taken as the terminal number estimate, and the channel information corresponding to each active codeword in the reference time slot is taken as the current cluster center. For each time slot other than the reference time slot, calculate the Euclidean distance between the channel information amplitude of each active codeword channel in the current time slot and the current cluster center, and construct a cost matrix based on the Euclidean distance; Based on the cost matrix, the Hungarian algorithm is used to assign the active codewords of the current time slot to the corresponding cluster centers to obtain the clustering results.

7. A passive access device for unmanned aerial vehicles (UAVs), characterized in that, include: The determination module is used to determine the data segment to be transmitted in each time slot sent by the passive terminal; The mapping module is used to determine the codeword of each time slot based on the data segment to be transmitted in each time slot according to the preset shared codebook tensor, and map the codeword of each time slot to the preset three-dimensional resource tensor to obtain the tensor signal of each time slot. The processing module is used to send the tensor signal of each time slot to a preset center point, and receive the preprocessed tensor of each time slot sent by the preset center point after performing tensor reshaping and domain transformation processing on the tensor signal of each time slot. The recovery module is used to perform channel tensor recovery processing on the preprocessed tensor of each time slot based on a preset cross-domain iterative detection strategy to obtain the recovery channel of each time slot, and to perform message concatenation based on the recovery channel of each time slot to obtain the recovery data of the passive terminal.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the passive access method for unmanned aerial vehicles as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the passive access method for unmanned aerial vehicles as described in any one of claims 1-6.

10. A computer program product, said computer program product storing a computer program, characterized in that, When executed by the processor, the program implements the passive access method for unmanned aerial vehicles as described in any one of claims 1-6.

Images