Unmanned aerial vehicle navigation method based on channel graph

Abstract

The invention belongs to the technical field of communication perception, and provides an unmanned aerial vehicle navigation method based on a channel graph, which comprises the following steps: firstly, deploying multiple base stations and multiple RISs in a GNSS denial environment to construct an ISAC system, collecting subcarrier domain CSI of each position by an unmanned aerial vehicle and converting the CSI into a time delay domain, then constructing a semi-supervised twin network framework based on UMAP, and finally constructing a semi-supervised twin network framework based on the UMAP; the method comprises the following steps: designing a mixed loss function fusing real coordinates and UMAP prior, carrying out two-dimensional DFT on time delay domain CSI to obtain an angle time delay domain, inputting network training to generate a channel graph, constructing three-dimensional simulation environment based on ray tracing to generate data, and training and evaluating the positioning precision of the channel graph. According to the method, RIS is used for expanding signal coverage, and the UMAP semi-supervised twin network is combined, so that the dependence on a sample label is remarkably reduced, and meter-scale positioning precision is realized in a dense city scene.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of communication perception, and particularly relates to a UAV navigation method based on a channel graph, a terminal device and a storage medium. BACKGROUND

[0002] UAV navigation is an important supporting technology for low-altitude economic development, and its core is to realize stable and accurate positioning in a complex environment. At present, various complementary technical solutions are mainly used in this field: a global satellite navigation system (GNSS) can provide a global absolute position reference; an inertial measurement unit (IMU) can measure acceleration and angular velocity in real time, and is used for short-time dead reckoning; a visual simultaneous localization and mapping (SLAM) technology perceives environmental geometric features through an airborne camera, constructs a map and simultaneously estimates the self-position; in addition, radio positioning methods using existing communication infrastructure are also concerned, mainly including model-driven methods based on channel parameter geometry calculation, and data-driven methods based on matching of pre-acquired signal fingerprints.

[0003] However, the current method has the problem that in a GNSS denial and non-line-of-sight propagation environment, existing radio positioning methods have high dependence on a large amount of labeled data, high deployment cost, and are difficult to realize high-precision navigation. SUMMARY

[0004] Embodiments of the application provide a UAV navigation method based on a channel graph, a terminal device and a storage medium, which can solve the problem that in a GNSS denial and non-line-of-sight propagation environment, existing radio positioning methods have high dependence on a large amount of labeled data, high deployment cost, and are difficult to realize high-precision navigation.

[0005] In a first aspect, the embodiments of the present application provide a channel graph based unmanned aerial vehicle navigation method, comprising: step 1, establishing a GNSS denial condition based unmanned aerial vehicle navigation ISAC system assisted by a multi-intelligent reflective surface RIS, using the RIS to expand the base station airspace coverage and regulate the wireless propagation environment to enhance the distinguishability of channel state information CSI; step 2, the unmanned aerial vehicle collects the subcarrier domain CSI corresponding to each sampling point position, and converts the subcarrier domain CSI into time delay domain CSI through discrete inverse Fourier transform; step 3, a semi-supervised twin neural network channel graph framework is constructed based on uniform manifold approximation and projection UMAP, which is used to process CSI data containing a first preset number of true value samples with positions and a second preset number of unlabeled samples, and a loss function based on true value coordinates is designed for the first preset number of true value sample CSI data, and a loss function based on UMAP prior information is designed for the second preset number of unlabeled sample CSI data; step 4, after the paired time delay domain CSI is processed by feature engineering, it is input into the twin neural network for training, the neural network parameters capable of mapping high-dimensional channel features to low-dimensional position embeddings are trained by jointly optimizing the loss function based on true value coordinates and the loss function based on UMAP prior information, and a channel graph reflecting the true position is output; step 5, a full-process simulation verification platform including channel modeling based on the ray tracing method, data acquisition and processing, network training and testing is built, the algorithm performance is systematically evaluated, and finally the semi-supervised twin neural network and the channel graph can be used for self-positioning and navigation of future unmanned aerial vehicles in the region.

[0006] In a possible implementation form of the first aspect, the above step 1 is specifically as follows:

[0007] The unmanned aerial vehicle receives pilot signals sent by multiple base stations at a preset fixed height with uniform time intervals, and the position of the unmanned aerial vehicle corresponding to each sampling time is called a sampling point, and the total number of sampling points is Q;

[0008] The unmanned aerial vehicle receives the OFDM signals sent by the multiple base stations in the scattering channel model with the RIS at each sampling point;

[0009] In the Q sampling points, the time of the qth sampling point is denoted as , the two-dimensional true position is denoted as , when the unmanned aerial vehicle is at the qth sampling point, the OFDM modulated pilot symbol sent by the kth base station is received , the number of subcarriers is , the subcarrier bandwidth is , and the received signal is denoted as , and the corresponding signal model is specifically as follows:

[0010]

[0011] wherein, is an additive white Gaussian noise with variance is a complex Gaussian distribution, is the CSI matrix corresponding to the qth sampling point, the kth base station and the lth subcarrier, and the corresponding expression is as follows:

[0012]

[0013] wherein, and represent the complex gain and delay of the lth path between the qth sampling point and the kth base station, respectively, is the frequency of the lth subcarrier, is the number of all communication paths between the qth sampling point and the kth base station, is a phase rotation factor, and j is an imaginary unit, and represent the steering vector of the UAV surface antenna array and the kth base station surface antenna array, respectively, and represent the azimuth angle and the elevation angle, respectively, ; the steering vector is specifically represented as follows:

[0014]

[0015] wherein, , , is the three-dimensional spatial position of each antenna element of the transmitting end or the receiving end relative to the geometric center of the array, is the number of antenna arrays of the transmitting end or the receiving end; is the wave vector, and is represented as:

[0016]

[0017] wherein, is the wavelength, includes a line of sight (LoS) path, an RIS path and a scattering path; wherein the complex gain of the RIS path can be represented as:

[0018]

[0019] wherein, and are the number of elements of the receiving end surface antenna x-axis and y-axis directions, g is the RIS unit coefficient, and represent the lth path between the qth sampling point and the kth base station.​​​ Path gain from each RIS unit to the drone and base station; and It is the first Normalized power radiation modes of a single RIS element in the receiving and reflecting directions; and The first The azimuth angle between the RIS element and the drone and base station and For the first The elevation angle between the RIS component and the drone and base station Indicates the first Phase shift of each RIS element; and Indicates the first The corresponding distances from each RIS component to the drone and base station.

[0020] Optionally, in another possible implementation of the first aspect, step 2 above is specifically as follows:

[0021] The q-th sampling point and the k-th base station are at the... The CSI on each subcarrier is obtained through a least-squares estimator, as shown below:

[0022]

[0023] in, For transmitting signals, H represents the conjugate transpose of the matrix;

[0024] use The point-based discrete inverse Fourier transform (IDFT) converts the CSI from the subcarrier domain to the time delay domain, as shown below:

[0025]

[0026] Where n represents the nth delay index or the nth propagation path; for all base stations, the high-dimensional delay domain CSI tensor corresponding to the sampling point, transmit / receive antenna, and multipath propagation. The statement is as follows:

[0027]

[0028] The number of delay dimensions is ,Right now , A set of complex tensors Number of base stations This refers to the number of antenna arrays on the base station side. This represents the number of antenna arrays for the drone.

[0029] Optionally, in another possible implementation manner of the first aspect, the design process of the loss function based on the UMAP prior information in step 3 is specifically as follows:

[0030] The set of sampling points with real landmark positions is denoted as , and the number of sampling points is The set of unmanned aerial vehicles without real positions is denoted as , and the number of unmanned aerial vehicles is For the sampling points in the set , the loss function of the training process is , and is defined as follows:

[0031]

[0032] wherein is the number of sampling points with real landmark positions, and represent the predicted positions and the true values of the channel map respectively;

[0033] For the unmanned aerial vehicle sampling points in the set , the expression of the loss function is as follows:

[0034]

[0035] wherein is the number of unlabeled sampling points, and the pair-wise dissimilarity and are the fusion dissimilarity and the UMAP prior dissimilarity, which are used for local and global constraints respectively, is a weight coefficient, which is used for measuring the contribution weight of the fusion dissimilarity and the UMAP prior dissimilarity;

[0036] wherein the fusion dissimilarity is composed of the temporal dissimilarity and the autocorrelation dissimilarity , and the calculation formula is as follows:

[0037]

[0038] wherein and are weight coefficients of the autocorrelation dissimilarity and the temporal dissimilarity respectively, and are the times of the i and j sampling points, is a time threshold; the expression of the temporal dissimilarity is as follows:

[0039]

[0040] Autocorrelation dissimilarity The definition is as follows:

[0041]

[0042]

[0043]

[0044]

[0045] in, For related functions, and Let i and j be the energy of CSI corresponding to sampling points i and j. It is the CSI matrix The The nth element represents the nth element received by the receiver. The channel state information measurement value between the b-th antenna and the transmitting antenna at the b-th antenna;

[0046] UMAP Prior Dissimilarity The expression is as follows:

[0047]

[0048] Wherein, the paired positions of the UMAP pre-dimensionality-reduced low-dimensional embeddings are denoted as follows: and .

[0049] Optionally, in another possible implementation of the first aspect, the pairwise delay domain CSIs in step 4 above are feature-engineered as follows:

[0050] high-dimensional time delay domain CSI tensor By separately analyzing the dimensions of the transmitting and receiving antennas and Point DFT transformation, converted to angle domain CSI, the specific DFT transformation is expressed as follows:

[0051]

[0052] The high-dimensional angular time delay domain CSI tensor is specifically represented as follows:

[0053] .

[0054] Optionally, in another possible implementation of the first aspect, step 5 above involves building a full-process simulation verification platform that includes channel modeling based on ray tracing methods, data acquisition and processing, network training and testing, as detailed below:

[0055] Based on the OpenStreetMap data, a dense urban environment area is imported into Blender software, and a coordinate system is created to define the material properties of the buildings in the dense urban environment area;

[0056] Based on the Sionna communication system simulation library, base stations, intelligent reflecting surfaces and unmanned aerial vehicles are added in the dense urban environment area, the transmitting and receiving antennas and RIS amplitude and phase are configured, the wireless environment is simulated through the ray tracing function, and the required channel parameters are generated;

[0057] An OFDM signal model is established according to the channel parameters to realize the collection and processing of CSI data;

[0058] The parameters of the twin neural network of the channel graph are trained, the accuracy of the channel graph for positioning is measured according to the visual effect and performance indicators, and the final implementation and verification of the algorithm are completed.

[0059] In the second aspect, the embodiments of the present application provide a terminal device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the channel graph-based unmanned aerial vehicle navigation method as described above when executing the computer program.

[0060] In the third aspect, the embodiments of the present application provide a computer readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the channel graph-based unmanned aerial vehicle navigation method as described above.

[0061] Beneficial effects: first establish a multi-intelligent reflective surface RIS assisted unmanned aerial vehicle navigation ISAC system under the condition of GNSS denial, use RIS to expand the base station airspace coverage and regulate the wireless propagation environment to enhance the distinguishability of channel state information CSI, then the unmanned aerial vehicle collects the subcarrier domain CSI corresponding to each sampling point position, and converts the subcarrier domain CSI into time delay domain CSI through discrete inverse Fourier transform, then a semi-supervised twin neural network channel graph framework is constructed based on uniform manifold approximation and projection UMAP, which is used to process CSI data containing a first preset number of true value samples with position and a second preset number of unlabeled samples, and a loss function based on true value coordinates is designed for the first preset number of true value sample CSI data with position, and a loss function based on UMAP prior information is designed for the second preset number of unlabeled sample CSI data, then the paired time delay domain CSI is processed through feature engineering and input into the twin neural network for training, the neural network parameters capable of mapping high-dimensional channel features to low-dimensional position embedding are trained by jointly optimizing the loss function based on true value coordinates and the loss function based on UMAP prior information, and the channel graph reflecting the real position is output, finally, a full-process simulation verification platform including channel modeling based on the ray tracing method, data acquisition and processing, network training and testing is built, the algorithm performance is systematically evaluated, and the finally trained semi-supervised twin neural network and channel graph can be used for self-positioning and navigation of future unmanned aerial vehicles in the region. The present application uses RIS to expand signal coverage, combines UMAP semi-supervised twin network, significantly reduces the dependence on sample labels, and realizes meter-level positioning accuracy in dense urban scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0063] Figure 1 is a flow diagram of a channel graph based unmanned aerial vehicle navigation method provided by an embodiment of the present application;

[0064] Figure 2 is a schematic diagram of a GNSS denial environment based RIS-ISAC unmanned aerial vehicle navigation system model provided by an embodiment of the present application;

[0065] Figure 3 is a semi-supervised twin neural network channel graph algorithm flowchart provided by an embodiment of the present application based on UMAP;

[0066] Figure 4is a full-process simulation verification platform schematic diagram provided by an embodiment of the present application;

[0067] Figure 5 is a channel graph test set for positioning in a dense scene and comparison result graph with other algorithms provided by an embodiment of the present application;

[0068] Figure 6 is a channel graph test set result graph for positioning in a sparse unmanned aerial vehicle channel scene in multiple propagation environments provided by an embodiment of the present application;

[0069] Figure 7 is a structure schematic diagram of a terminal device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0070] In the following description, specific details are set forth, such as particular system configurations, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, persons skilled in the art will understand that the present application can be practiced without these specific details. In other instances, well-known systems, structures, circuits, and techniques have not been shown in detail in order not to obscure the understanding of this application.

[0071] It should be understood that the term "comprises" as used in the specification and the appended claims indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0072] It should also be understood that the term "and / or" as used herein refers to any combination of associated listed items, as well as all possible combinations thereof, and includes these combinations.

[0073] As used in the specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "if it is determined" or "if [the described condition or event] is detected" can be interpreted as meaning "upon determining" or "in response to determining" or "upon detecting [the described condition or event]" or "in response to detecting [the described condition or event]" depending on the context.

[0074] In addition, in the description of the present application and the appended claims, the terms "first", "second", "third", etc. are only used for differentiation in description, and cannot be understood as indicating or implying relative importance.

[0075] Reference throughout this application to "one embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearances of the phrase "in one embodiment" or "in some embodiments" in various places throughout this specification are not necessarily all referring to the same embodiment, however, it is contemplated that the features, structures, or characteristics of one embodiment can be combined with those of other embodiments. The terms "including", "comprising", "having" and the like are meant to be inclusive and mean that there can be additional

[0076] A channel map based UAV navigation method, a terminal device and a storage medium provided by the present application are described in detail below with reference to the accompanying drawings.

[0077] Figure 1 A flowchart of a channel map based UAV navigation method provided by an embodiment of the present application is shown.

[0078] As Figure 1 shown, the channel map based UAV navigation method includes the following steps:

[0079] Step 1, establish a GNSS denial condition based multi-intelligent reflective surface RIS assisted UAV navigation ISAC system, use RIS to expand base station airspace coverage and regulate wireless propagation environment to enhance the discrimination degree of channel state information CSI;

[0080] It should be noted that, as Figure 2 shown, the UAV receives pilot signals sent by multiple base stations at a preset fixed height with uniform time intervals, and the UAV position corresponding to each sampling time is called a sampling point; the UAV receives an orthogonal frequency division multiplexing OFDM signal under a scattering channel model with RIS at each sampling point, and the total number of sampling points is Q. RIS can enhance airspace coverage and improve the discrimination degree of CSI.

[0081] Further, in the embodiment of the present application, the above step 1 includes:

[0082] The UAV receives pilot signals sent by multiple base stations at a preset fixed height with uniform time intervals, and the UAV position corresponding to each sampling time is called a sampling point, and the total number of sampling points is Q;

[0083] The UAV receives an orthogonal frequency division multiplexing OFDM signal sent by multiple base stations under a scattering channel model with RIS at each sampling point;

[0084] In the Q sampling points, the q-th sampling point time is denoted as , and the two-dimensional real position is denoted as When the UAV is at the qth sampling point, the OFDM modulated pilot symbol transmitted by the kth base station is received , the number of subcarriers is , the subcarrier bandwidth is , the received signal is denoted as , and the corresponding signal model is as follows:

[0085]

[0086] wherein, is an additive white Gaussian noise with a variance of , is a complex Gaussian distribution, is the CSI matrix corresponding to the qth sampling point, the kth base station and the th subcarrier, and the corresponding expression is as follows:

[0087]

[0088] wherein, and represent the complex gain and delay of the lth path between the qth sampling point and the kth base station, is the frequency of the th subcarrier, is the number of all communication paths between the qth sampling point and the kth base station, is a phase rotation factor, and j is an imaginary unit, and represent the steering vectors of the UAV surface antenna array and the kth base station surface antenna array, and represent the azimuth angle and the elevation angle, ; the steering vectors are specifically represented as follows:

[0089]

[0090] wherein, , , is the three-dimensional spatial position of each antenna element of the transmitting end or the receiving end relative to the geometric center of the array, is the number of antenna arrays of the transmitting end or the receiving end; is the wave vector, and is represented as:

[0091]

[0092] wherein, is the wavelength, includes a line of sight LoS path, an RIS path and a scattering path; wherein the complex gain of the RIS path can be represented as: ​

[0093]

[0094] wherein, and are the number of elements of the receiving end-face antenna in the x-axis and y-axis directions, respectively, g is the RIS element coefficient, and denote the path gain from the th RIS element to the UAV and the base station, respectively; and are the normalized power radiation patterns of the th RIS element in the receiving direction and the reflecting direction, respectively; and are the azimuth angles of the th RIS element to the UAV and the base station, respectively, and are the elevation angles of the th RIS element to the UAV and the base station, respectively, denotes the phase shift of the th RIS element; and denote the corresponding distances of the th RIS element to the UAV and the base station, respectively.

[0095] Step 2, the UAV collects the subcarrier domain CSI corresponding to each sampling point position, and converts the subcarrier domain CSI into time delay domain CSI through a discrete inverse Fourier transform;

[0096] Further, in the embodiments of the present application, the above-mentioned step 2 comprises:

[0097] The CSI of the th sampling point and the th base station on the th subcarrier is obtained through a least square estimator, and is specifically represented as follows:

[0098]

[0099] wherein, is the transmitting signal, and H denotes the conjugate transpose of the matrix;

[0100] It should be noted that, compared with the subcarrier domain CSI, the time delay domain CSI more directly reveals the multipath structure of the channel, each delay tap corresponding to a propagation path. Its typical sparse characteristics reduce the data dimension, memory occupation and computational complexity. In addition, it helps to extract distance-related features that are beneficial to the positioning of the UAV.

[0101] The point-discrete inverse Fourier transform (IDFT) converts the CSI from the subcarrier domain to the delay domain, and is expressed as follows:

[0102]

[0103] where n represents the nth delay index or the nth propagation path; for all base stations, the high-dimensional delay domain CSI tensor corresponding to the sampling points, the transmit / receive antennas and the multipath propagation is expressed as follows:

[0104]

[0105] The number of delay dimensions of is , is a set of complex tensors, is the number of base stations, is the number of base station side antenna arrays, is the number of UAV antenna arrays.

[0106] Step 3, based on the unified manifold approximation and projection (UMAP), a semi-supervised twin neural network channel graph framework is constructed for processing CSI data containing a first preset number of true value samples with positions and a second preset number of unlabeled samples, and a loss function based on true value coordinates is designed for the first preset number of true value sample CSI data, and a loss function based on UMAP prior information is designed for the second preset number of unlabeled sample CSI data;

[0107] Further, in the embodiments of the present application, the design process of the loss function based on UMAP prior information in step 3 above is as follows:

[0108] It should be noted that under the condition of GNSS denial, considering that most of the UAV sampling points do not have real position labels, however, due to the existence of landmark buildings, a small part of the coordinates of the UAV sampling points can be collected. The algorithm designs a loss function according to whether the sampling points have real position information.

[0109] Let the set of sampling points with landmark real positions be , the number of which is , and the set of UAVs without real positions be , the number of which is ; for the sampling points in the set , the loss function of the training process is , which is defined as follows:

[0110]

[0111] wherein, the number of sampling points of the real position of the landmark, and respectively represent the channel map prediction position and the true value position;

[0112] It should be noted that for the unmanned aerial vehicle sampling points in the set In the absence of accurate position labels, the dissimilarity measure of the pair of points is directly derived from the intrinsic characteristics of the CSI data, so as to maintain the topological relationship between the sampling points. This dissimilarity calculation must satisfy two key constraint conditions: maintaining local neighborhood consistency to capture fine-grained spatial relationships and maintaining global structural integrity to prevent distortion-induced positioning errors.

[0113] For the unmanned aerial vehicle sampling points in the set The expression of the loss function is as follows:

[0114]

[0115] wherein, is the number of unlabeled sampling points, and the dissimilarity of the pair and is the fusion dissimilarity and the UMAP prior dissimilarity, respectively, for local and global constraints, is the weight coefficient, used to measure the contribution weight of the fusion dissimilarity and the UMAP prior dissimilarity;

[0116] It should be noted that UMAP is a manifold learning-based dimensionality reduction algorithm that can maintain global structure while considering local features during dimensionality reduction, and has the advantages of fast calculation speed and easy visualization, suitable for dimensionality reduction of high-dimensional nonlinear data. Therefore, the pair-wise relative distance obtained by UMAP pre-reduction is the global compensation term in the loss function.

[0117] wherein the fusion dissimilarity is composed of the temporal dissimilarity and the autocorrelation dissimilarity The calculation formula is as follows:

[0118]

[0119] wherein, and are the weight coefficients of the autocorrelation dissimilarity and the temporal dissimilarity, respectively, and are the times of the i and j sampling points, is the time threshold; the expression of the temporal dissimilarity is as follows:

[0120]

[0121] Self-correlation dissimilarity is defined as follows:

[0122]

[0123]

[0124]

[0125]

[0126] wherein, is a correlation function, and is the energy of the CSI corresponding to the i and j sampling points, is the element of the CSI matrix , representing the channel state information measurement value between the i-th receiving antenna and the b-th transmitting antenna; UMAP prior dissimilarity is expressed as follows:

[0127]

[0128] wherein, the pair-wise positions of the UMAP pre-reduced dimension low-dimensional embedding are respectively denoted as and

[0129] . Step 4, after the pair-wise time-delay domain CSI is processed by feature engineering, it is input into the twin neural network for training, by jointly optimizing the loss function based on the true value coordinates and the loss function based on the UMAP prior information, the neural network parameters capable of mapping the high-dimensional channel features to the low-dimensional position embedding are trained, and the channel graph reflecting the real position is output;

[0130] Further, in the embodiments of the present application, as shown in , the pair-wise time-delay domain CSI in the above step 4 is processed by feature engineering, specifically as follows:

[0131] Figure 3 The high-dimensional time-delay domain CSI tensor is converted into an angle domain CSI by performing DFT transformation on the transmitting and receiving antenna dimensions respectively, and the specific DFT transformation is represented as:

[0132] The high-dimensional angle time-delay domain CSI tensor is specifically represented as:

[0133]

[0134] ​​​​

[0135] .

[0136] In the embodiments of the present application, through the above processing, the high-dimensional time delay domain CSI tensor in the antenna domain is converted to the angle domain, which can better assist the neural network by capturing the angle characteristics, thereby realizing high-precision positioning.

[0137] The pair of high-dimensional angle-time delay domain CSI is input to the input end of the twin neural network, and two loss functions are respectively minimized and optimized according to whether the input data has a label, to obtain neural network parameters and a channel graph reflecting the real position.

[0138] Step 5, a full-process simulation verification platform including channel modeling based on the ray tracing method, data collection and processing, network training and testing is built, the performance of the algorithm is systematically evaluated, and finally the semi-supervised twin neural network and the channel graph trained can be used for self-positioning and navigation of future unmanned aerial vehicles in the region.

[0139] Further, in the embodiments of the present application, as shown in Figure 4 the full-process simulation verification platform including channel modeling based on the ray tracing method, data collection and processing, network training and testing in step 5 is built, and the specific steps are as follows:

[0140] Based on the OpenStreetMap data, a dense urban environment area is imported into the Blender software, and a coordinate system is created to define the material properties of the buildings in the dense urban environment area.

[0141] Based on the Sionna communication system simulation library, base stations, intelligent reflecting surfaces and unmanned aerial vehicles are added in the dense urban environment area, the transmitting and receiving antennas and the RIS amplitude and phase are configured, the wireless environment is simulated through the ray tracing function, and the required channel parameters are generated.

[0142] An OFDM signal model is established according to the channel parameters to realize the collection and processing of CSI data.

[0143] The twin neural network parameters of the channel graph are trained, and the accuracy of the channel graph for positioning is measured according to the visualization effect and performance indicators, and the final implementation and verification of the algorithm are completed.

[0144] In the embodiments of the present application, the specific simulation parameter settings are shown in Table 1; the algorithm effect comparison and verification visualization results in the unmanned aerial vehicle dense sampling scene are shown in Figure 5 , the channel graph visualization results in the unmanned aerial vehicle channel scene are shown in Figure 6 ; the positioning accuracy numerical simulation results in the dense scene are shown in Table 2, and the channel graph positioning accuracy numerical results in the unmanned aerial vehicle channel scene are shown in Table 3.

[0145] Table 1

[0146]

[0147] Table 2

[0148]

[0149] Table 3

[0150]

[0151] Figure 5 Table 1 and Table 2 respectively reflect the multiple advantages of the present application from the perspective of visualization and numerically: (1) Only 10-20% of the labeled data is needed to reconstruct almost distortion-free channel graphs. Compared with the pure UMAP method and the geodesic-based twin network, the proposed method can more accurately restore the neighborhood relationship and maintain the global structure, and the generated channel graph has higher quality. (2) The positioning accuracy reaches the level comparable to the fingerprint positioning method, but the amount of label required is greatly reduced. When using 100% labeled real data, the present application can achieve an average absolute error of 2.67 meters, which is better than WKNN (5.49 meters) and CNN fingerprint positioning method (2.91 meters); it is worth noting that in the case of only using 20% labeled data, the present application can achieve a positioning error of 3.66 meters, which is significantly better than WKNN (10.4 meters) and CNN fingerprint positioning method (8.89 meters) using 80% labeled samples. This means that while reducing the amount of real data collection by 4 times, the positioning accuracy is more than 2 times higher than the traditional fingerprint method.

[0152] Figure 6 Table 3 respectively reflects the influence of RIS and line-of-sight conditions in the scenario of unmanned aerial vehicles flying along the channel from the perspective of visualization and numerically. The results show that the algorithm still maintains a certain robustness and achieves stable positioning performance under the condition of a significant decrease in the size of the data set: it reaches an average positioning error of 1.51 meters in the LoS condition and an average positioning error of 2.71 meters in the NLoS condition. In the LoS condition, the channel graph quality and the unmanned aerial vehicle positioning accuracy are compared when RIS exists and does not exist. When lacking RIS, not only does the channel graph quality deteriorate sharply (the contour of the figure appears distorted), but the positioning error also rises to 3.75 meters. This proves that the model equipped with RIS can collect higher quality CSI data and achieve better discrimination at different positions.

[0153] The application provides a UAV navigation method based on a channel graph, which comprises the following steps: first, establishing a UAV navigation ISAC system assisted by a multi-intelligent reflective surface RIS under GNSS denial conditions, expanding the base station space coverage and regulating the wireless propagation environment by using the RIS, and enhancing the distinguishability of channel state information CSI; then, collecting, by the UAV, subcarrier domain CSI corresponding to the position of each sampling point, and converting the subcarrier domain CSI into time delay domain CSI through discrete inverse Fourier transform; next, constructing a semi-supervised twin neural network channel graph framework based on uniform manifold approximation and projection UMAP, for processing CSI data containing a first preset number of true value samples with positions and CSI data containing a second preset number of unlabeled samples, designing a loss function based on true value coordinates for the first preset number of true value sample CSI data, and designing a loss function based on UMAP prior information for the second preset number of unlabeled sample CSI data; then, inputting the pairs of time delay domain CSI into the twin neural network after feature engineering processing, training the twin neural network through joint optimization of the loss function based on the true value coordinates and the loss function based on the UMAP prior information, obtaining neural network parameters capable of mapping high-dimensional channel features to low-dimensional position embeddings, and outputting a channel graph reflecting the real position; finally, building a full-process simulation verification platform including channel modeling based on a ray tracing method, data acquisition and processing, network training and testing, systematically evaluating the algorithm performance, and finally using the semi-supervised twin neural network and the channel graph to realize self-positioning and navigation of the UAV in the region. The application uses the RIS to expand signal coverage, and combines the UMAP semi-supervised twin network, thereby significantly reducing the dependence on sample labels and realizing meter-level positioning accuracy in a dense urban scenario.

[0154] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the application.

[0155] In order to realize the above-mentioned embodiments, the application further provides a terminal device.

[0156] Figure 7 The structural schematic diagram of the terminal device of one embodiment of the application.

[0157] As shown in Figure 7 , the terminal device 200 comprises:

[0158] a memory 210 and at least one processor 220, a bus 230 connecting different components (including the memory 210 and the processor 220), and the memory 210 stores a computer program, which realizes the UAV navigation method based on the channel graph when the processor 220 executes the program.

[0159] Bus 230 generally represents what in the art will be broadly understood as a bus structure for a communication architecture. It can represent one or more of several types of bus structures including memory buses or memory controllers, peripheral buses, graphics acceleration buses (e.g., AGP bus) and a local bus using various bus architectures (e.g., Industrial Standard Architecture (ISA), Micro Channel Architecture (MCA), Enhanced ISA (EISA), Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus).

[0160] Terminal device 200 typically includes a variety of computer system readable media. These media can be any available media that is accessible by terminal device 200 and includes both volatile and non-volatile media, removable and non-removable media.

[0161] Memory 210 also can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 240 and / or cache memory 250. Terminal device 200 can further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 260 can be provided for reading from and writing to a non-removable, non-volatile magnetic media (e.g., a "hard drive"). Figure 7 not shown, is typically provided as residual storage across terminal device 200, and can be used for storing data that is both received as well as data that is generated by the processor 220. Although Figure 7 not shown, is typically provided as residual storage across terminal device 200, and can be used for storing data that is both received as well as data that is generated by the processor 220. Although

[0162] Program / utility 280, having a set of programs / modules 270, can be stored in, for example, memory 210 by way of example, such programs includes, but is not limited to, an operating system, one or more applications, other program modules 270, and program data, each of which or a combination can include implementation of a networking environment. Program modules 270 generally carry out the functions and / or methodologies of embodiments of the application as described herein.

[0163] Terminal device 200 can also be in communication with one or more external devices 290 such as a keyboard, a pointing device, a display 291, etc.; one or more devices that enable a user to interact with terminal device 200; and / or any devices (e.g., network card, modem, etc.) that enable terminal device 200 to communicate with one or more other computing devices. Such communication can be facilitated by an Input / Output (I / O) interface 292. Still yet, terminal device 200 can be in communication with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or the Internet) through a network adapter 293. As depicted, network adapter 293 communicates with the other components of terminal device 200 through bus 230. It should be appreciated that although not shown, other hardware and / or software modules could be used in conjunction with terminal device 200. Such modules include, but are not limited to, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems, etc.

[0164] Processor 220 performs various function applications and data processing by running programs stored in memory 210.

[0165] It should be noted that the implementation process and technical principles of the terminal device of the embodiment are referred to the aforementioned explanation and description of the method for unmanned aerial vehicle navigation based on channel map, and will not be repeated here.

[0166] The embodiment of the present application further provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the steps in each method embodiment.

[0167] The embodiment of the present application provides a computer program product, when the computer program product is run on a terminal device, the terminal device is enabled to realize the steps in each method embodiment.

[0168] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes in the above-mentioned embodiment methods through a computer program to instruct relevant hardware to complete, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium can at least include any entity or device capable of carrying the computer program code to the photographing device / terminal equipment, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunication signal.

[0169] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0170] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be realized by electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0171] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal equipment and method can be implemented in other ways. For example, the above-described apparatus / terminal equipment embodiments are merely schematic, and the division of the modules or units is merely a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0172] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may also be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0173] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A UAV navigation method based on channel graphs, characterized in that, Includes the following steps: Step 1: Establish an ISAC system for UAV navigation assisted by multiple intelligent reflectors (RIS) under GNSS denial conditions. RIS is used to extend the airspace coverage of the base station and regulate the wireless propagation environment to enhance the distinguishability of Channel State Information (CSI). Step 2: The UAV collects the subcarrier domain CSI corresponding to each sampling point location, and converts the subcarrier domain CSI into the time delay domain CSI through discrete inverse Fourier transform; Step 3: Construct a semi-supervised twin neural network channel graph framework based on Unified Manifold Approximation and Projection UMAP to process CSI data containing a first preset number of CSI data with location truth samples and a second preset number of CSI data without labels. Design a loss function based on truth coordinates for the first preset number of CSI data with location truth samples and a loss function based on UMAP prior information for the second preset number of CSI data without labels. Step 4: After performing feature engineering on the paired delay domain CSIs, input them into the Siamese neural network for training. By jointly optimizing the loss function based on the ground truth coordinates and the loss function based on UMAP prior information, the neural network parameters that can map high-dimensional channel features to low-dimensional location embeddings are trained, and a channel map reflecting the real location is output. Step 5: Build a full-process simulation verification platform that includes channel modeling, data acquisition and processing, network training and testing based on ray tracing methods, systematically evaluate the algorithm performance, and finally train a semi-supervised twin neural network and channel map that can be used for the self-localization and navigation of future UAVs in this area.

2. The UAV navigation method based on channel map according to claim 1, characterized in that, Step 1 is as follows: The UAV receives pilot signals from multiple base stations at a preset fixed altitude at uniform time intervals. The position of the UAV at each sampling moment is called a sampling point, and the total number of sampling points is Q. The UAV receives orthogonal frequency division multiplexing (OFDM) signals transmitted by multiple base stations under a scattering channel model with RIS at each sampling point; In the Q sampling points, the time of the q-th sampling point is denoted as . The two-dimensional real position is denoted as When the UAV receives the OFDM modulated pilot symbol sent by the k-th base station at the q-th sampling point. The number of subcarriers is The subcarrier bandwidth is The received signal is denoted as The corresponding signal model is as follows: ； in, The variance is Additive white Gaussian noise, It follows a complex Gaussian distribution. For the q-th sampling point, the k-th base station, and the... The CSI matrix corresponding to each subcarrier is expressed as follows: ； in, and Let these represent the complex gain and delay of the l-th path between the q-th sampling point and the k-th base station, respectively. It is the first The frequency of each subcarrier This represents the total number of communication paths between the q-th sampling point and the k-th base station. Here, j is the phase rotation factor, and j is the imaginary unit. and These represent the steering vectors of the UAV surface antenna array and the k-th base station surface antenna array, respectively. and These represent azimuth and elevation angles, respectively. The guide vector is specifically represented as follows: ； in, , , This refers to the three-dimensional spatial position of each antenna element at the transmitting or receiving end relative to the geometric center of the array. This refers to the number of antenna arrays at the transmitting or receiving end. The wave vector is represented as: ； in, For wavelength, the It includes a line-of-sight (LoS) path, a RIS path, and a scattering path; the complex gain of the RIS path can be expressed as: ； in, and These represent the number of array elements along the x-axis and y-axis of the receiving end-face antenna, respectively. g represents the RIS unit coefficient. and They represent the first Path gain from each RIS unit to the drone and base station; and It is the first Normalized power radiation modes of a single RIS element in the receiving and reflecting directions; and The first The azimuth angle between the RIS element and the drone and base station and For the first The elevation angle between the RIS component and the drone and base station Indicates the first Phase shift of each RIS element; and Indicates the first The corresponding distances from each RIS component to the drone and base station.

3. The UAV navigation method based on channel maps according to claim 2, characterized in that, Step 2 is described in detail below: The q-th sampling point and the k-th base station are at the... The CSI on each subcarrier is obtained through a least-squares estimator, as shown below: ； in, For transmitting signals, H represents the conjugate transpose of the matrix; use The point-based discrete inverse Fourier transform (IDFT) converts the CSI from the subcarrier domain to the time delay domain, as shown below: ； Where n represents the nth delay index or the nth propagation path; for all base stations, the high-dimensional delay domain CSI tensor corresponding to the sampling point, transmit / receive antenna, and multipath propagation. The expression is as follows: ； The number of delay dimensions is ,Right now , A set of complex tensors Number of base stations This refers to the number of antenna arrays on the base station side. This represents the number of antenna arrays for the drone.

4. The UAV navigation method based on channel map according to claim 3, characterized in that, The design process of the loss function based on UMAP prior information in step 3 is as follows: Let the set of sampling points with actual landmark locations be denoted as The quantity is A collection of drones that do not have a real location is denoted as The quantity is For sets The sampling points in the dataset have a loss function for the training process as follows: The definition is as follows: ； in, This represents the number of sampling points with actual landmark locations. and These represent the channel diagram prediction position and the true value position, respectively. For sets The expression for the loss function of the UAV sampling points is as follows: ； in, The number of unlabeled sampling points, and the pairwise dissimilarity. and To integrate dissimilarity and UMAP prior dissimilarity, local and global constraints are applied separately. , which is a weighting coefficient used to measure the contribution weight of fusion dissimilarity and UMAP prior dissimilarity; Among them, fusion dissimilarity By time dissimilarity Autocorrelation dissimilarity The formula for calculating the fusion composition is as follows: ； in, and These are the weighting coefficients for autocorrelation dissimilarity and temporal dissimilarity, respectively. and For sampling points i and j, Time threshold; temporal dissimilarity The specific expression is as follows: ； Autocorrelation dissimilarity The definition is as follows: ； ； ； ； in, For related functions, and Let i and j be the energy of CSI corresponding to sampling points i and j. It is the CSI matrix The The nth element represents the nth element received by the receiver. The channel state information measurement value between the b-th antenna and the transmitting antenna at the b-th antenna; UMAP prior dissimilarity The expression is as follows: ； Wherein, the paired positions of the UMAP pre-dimensionality-reduced low-dimensional embeddings are denoted as follows: and .

5. The UAV navigation method based on channel map according to claim 4, characterized in that, The feature engineering process performed on the paired delay domain CSIs in step 4 is as follows: high-dimensional time delay domain CSI tensor By separately analyzing the dimensions of the transmitting and receiving antennas and Point DFT transformation, converted to angle domain CSI, the specific DFT transformation is expressed as follows: ； The high-dimensional angular time delay domain CSI tensor is specifically represented as follows: 。 6. The UAV navigation method based on channel map according to claim 5, characterized in that, Step 5 involves building a full-process simulation and verification platform that includes channel modeling, data acquisition and processing, network training and testing based on ray tracing methods, as detailed below: Based on OpenStreetMap data, a dense urban environment area was imported into Blender software, and a coordinate system was created to define the material properties of buildings in the dense urban environment area. Based on the Sionna communication system simulation library, base stations, smart reflectors, and drones are added in a dense urban environment. Transceiver antennas and RIS amplitude and phase are configured. The wireless environment is simulated through ray tracing to generate the required channel parameters. An OFDM signal model is established based on the channel parameters to realize the collection and processing of CSI data; The parameters of the twin neural network for the channel graph are trained, and the accuracy of the channel graph for positioning is measured based on visualization effects and performance indicators, thus completing the final implementation and verification of the algorithm.

7. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 6.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images