Relative positioning method based on channel splicing ranging in urban canyon environment

Abstract

The application relates to a relative positioning method based on channel splicing ranging in an urban canyon environment, and belongs to the technical field of wireless positioning. The method adopts a three-node cooperative positioning structure, including a fixed anchor point, a passenger terminal and a driver terminal. The fixed anchor point is arranged at a fixed position with a known coordinate. The passenger terminal remains relatively static during the positioning process. The driver terminal is arranged on a pickup vehicle and can obtain the position of the driver terminal. The method realizes wireless precise ranging by splicing multiple channels based on frequency hopping communication, obtains distance data between the fixed anchor point, the passenger terminal and the driver terminal, and combines the position information of the driver terminal to perform high-precision estimation on the relative position of the passenger terminal by using an extended Kalman filter or an unscented Kalman filter. The application does not depend on high-precision GNSS positioning results, and can realize high-precision positioning of less than 2 meters in an urban canyon scene.

Description

Technical Field

[0001] This invention belongs to the field of wireless positioning technology, and relates to positioning in urban canyon environments, specifically to a relative positioning method based on channel stitching ranging in urban canyon environments. Background Technology

[0002] In recent years, with the proliferation of high-rise buildings in cities, creating so-called "urban canyons," satellite navigation signals are often severely obstructed and affected by multipath reflections in urban streets. This leads to a significant reduction in the positioning accuracy of GNSS (Global Navigation Satellite System, such as GPS and BeiDou), often resulting in errors of tens of meters. For example, in ride-hailing scenarios, passengers typically upload their location information via mobile devices, but in densely built-up areas, this location may be far from the passenger's actual location, making it difficult for drivers to locate passengers promptly and accurately. To improve positioning accuracy, current technologies often use a combination of GNSS and IMU (Inertial Measurement Unit) for vehicle positioning to maintain position estimation even during short periods of poor satellite signal coverage. However, in urban canyon environments, a simple GNSS / IMU combination may still accumulate errors due to prolonged satellite signal obstruction, making it difficult to guarantee accurate passenger location. In addition, some technologies attempt to utilize local wireless signals for assisted positioning, such as measuring the relative distance between the vehicle and passenger via Bluetooth or Wi-Fi. However, these methods have limited accuracy in open environments and are not ideal for the dynamic ride-hailing process. Therefore, there is an urgent need for a new high-precision positioning technology that can achieve precise relative positioning between vehicles and passengers in complex urban canyon environments where satellite signals are limited, helping drivers to reach passengers smoothly and improving the ride-hailing experience. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a relative positioning method based on channel stitching ranging in urban canyon environments, which can achieve high-precision static-dynamic target relative positioning in urban canyon scenarios without relying on high-precision GNSS positioning results, and help dynamic targets quickly and accurately determine the position of static targets relative to themselves.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A relative positioning method based on channel stitching ranging in an urban canyon environment, the method comprising:

[0006] A three-node collaborative positioning structure is adopted, with fixed anchor points, passenger terminals and driver terminals set up to obtain the position and speed information of each node at the current moment;

[0007] Frequency hopping communication is performed using OFDM wireless signals between fixed anchor points, passenger terminals, and driver terminals to collect channel status information of multiple frequency bands;

[0008] A sparse reconstruction algorithm is used to identify a stable path in each frequency band. The stable path of one frequency band is selected as a reference, and the channel state information of other frequency bands is calibrated in terms of time and phase to initially eliminate frequency hopping channel errors.

[0009] The channel state information of each frequency band after correction is spliced ​​and integrated to construct an equivalent broadband channel frequency response; then the spliced ​​channel is regarded as composed of several frequency band blocks with unknown phase shift, and the relative impulse response of the spliced ​​channel is reconstructed through sparse inversion.

[0010] Two-way channel state information is obtained through two-way communication between the fixed anchor point, passenger terminal, and driver terminal. The two-way channel frequency response of each channel center frequency subcarrier is multiplied to obtain the squared CSI value, which is used as the channel frequency response sample for each frequency band. Based on the actual channel frequency response value and relative channel frequency response value of each frequency band channel frequency response sample, an error function is constructed to solve for the error parameters. Then, the frequency hopping channel error is eliminated based on the error parameters to obtain error-free spliced ​​channel state information.

[0011] The flight time of the signal is estimated based on error-free spliced ​​channel state information to obtain the distance measurements between each pair of the fixed anchor point, passenger terminal and driver terminal, wherein distance data is periodically collected at a predetermined frequency.

[0012] The position of the driver terminal at the next moment is predicted using a discrete linear motion model. The measured distances between each pair of the fixed anchor point, passenger terminal, and driver terminal are compared with the theoretical distances between each pair of the fixed anchor point, passenger terminal, and driver terminal calculated based on the position prediction to obtain the observation residuals. Based on the observation residuals, Kalman filtering is used to correct the predicted position of the passenger terminal.

[0013] Repeat the above process until the estimated distance between the driver terminal and the passenger terminal is less than a preset threshold, and maintain this for multiple measurement cycles, then determine that the driver terminal has reached the location of the passenger terminal.

[0014] Furthermore, a sparse reconstruction algorithm is used to identify a stable path in each frequency band. For any frequency band, its channel frequency response is represented as a linear combination of multipath parameters, and an orthogonal matching pursuit algorithm is used to find the best matching path in the overcomplete time delay dictionary. The path with the maximum energy is extracted through finite iterations, which is the stable path of that frequency band.

[0015] Furthermore, after identifying a stable path for each frequency band, a stable path for any frequency band w0 is selected as a reference, and the time delay difference between the stable paths of other frequency bands w and the stable paths of frequency band w0 is calculated. The estimated time delay for the stable path in frequency band w0. δ is the estimated time delay of the stable path in frequency band w. w This refers to the phase shift in frequency band w caused by sampling frequency deviation and time synchronization error. The phase shift in frequency band w0 caused by sampling frequency deviation and time synchronization error;

[0016] Multiply the spectrum of all subcarriers in frequency band w by a linear phase factor exp(+j2πfΔt) to advance the stable path delay of frequency band w by Δt, thereby aligning it with the stable path of frequency band w0, where f is the frequency of the subcarrier; simultaneously, apply a global phase rotation factor to the CSI composed of the channel frequency response of all subcarriers in frequency band w. Eliminate the initial phase difference between the stable path of frequency band w and the stable path of frequency band w0. The initial phase of the stable path for frequency band w.

[0017] Furthermore, the corrected channel state information of each frequency band is spliced ​​and integrated to construct an equivalent broadband channel frequency response, that is, to form a frequency vector f = [f] containing all frequency band subcarriers. (1) ,f (2) ,…,f (WM) ], f (WM) Let M be the frequency vector of the Mth subcarrier in the Wth frequency band, and the corresponding concatenated CSI vector. in This represents a vector of M corrected CSI values ​​over frequency band W;

[0018] The spliced ​​channel is considered as consisting of several frequency band blocks with unknown phase shifts. The relative impulse response of the spliced ​​channel is reconstructed using a sparse inversion method, and the spliced ​​CSI vector is modeled as follows:

[0019]

[0020] In the formula, F is the Fourier coefficient matrix constructed based on the frequency vector f, r is the sparse impulse response coefficient vector to be determined, and n is the noise term; the sparse solution of r is obtained by solving the following optimization problem:

[0021]

[0022] In the formula, β is the regularization parameter that controls sparsity, and ||·|2 and ||·|0 represent the L2 and L0 norms, respectively. The soft threshold iterative descent method is used to solve this optimization problem.

[0023] Furthermore, the soft threshold iterative descent method is used to solve the optimization problem. This includes: calculating the gradient between the residual and the Fourier coefficient matrix F at iteration t. * denotes the conjugate transpose of the matrix, r (t) Let r be the sparse impulse response coefficient vector of the t-th iteration; determine the step size s through line search; update the estimated r. (t+1) =r (t) +εg;for r (t+1) A soft thresholding process is performed, setting coefficients smaller than the threshold βε to zero to maintain the sparsity of the results. Through iterative iteration, the process stops when the relative error or residual converges to a predetermined threshold, yielding the relative impulse response of the spliced ​​channel.

[0024] Furthermore, regarding the relative impulse response Perform FFT to obtain the spectrum Then, the bidirectional channel frequency response is used to analyze the spectrum. Calibration includes:

[0025] Bidirectional Channel Frequency Response (CSI) between the fixed anchor point, passenger terminal, and driver terminal is obtained through a fast two-way communication mechanism. The bidirectional channel frequency response of each channel center frequency subcarrier is multiplied to obtain the squared CSI value, which is then used as a channel frequency response sample for each frequency band. The actual channel frequency response value and the relative channel frequency response value of each frequency band sample are compared to construct an error function. Solving this error function yields the global phase offset estimate. and time deviation parameter estimates For each subcarrier frequency point f m,w Multiply by the corrected phase Obtain the calibrated true channel frequency response H m,w The actual CSI spectrum H, after removing frequency hopping channel errors, is obtained by combining the real channel frequency responses of each subcarrier, where m = 1, 2, ..., M, M is the number of subcarriers in each frequency band, and w = 1, 2, ..., W, W is the number of frequency bands.

[0026] Furthermore, the flight time of the signal is estimated based on the spectrum H to obtain the distance measurements between each pair of the fixed anchor point, passenger terminal, and driver terminal;

[0027] The driver terminal's position at the next moment is predicted using a discrete linear motion model, which is expressed as follows:

[0028] x k∣k-1 =Fx k-1∣k-1 +w k-1 ,

[0029] In the formula, w k-1 Let G be the process noise, G be the state transition matrix, and x be the process noise. k-1∣k-1 To determine the system state x k-1 The posterior estimate, where Q is the process noise covariance matrix;

[0030] The measured distances between each pair of the fixed anchor point, passenger terminal, and driver terminal are compared with the theoretical distances between each pair of the fixed anchor point, passenger terminal, and driver terminal calculated based on the position prediction values ​​to obtain the observation residuals. Based on the observation residuals, the position prediction values ​​of the passenger terminal are corrected using extended Kalman filtering or unscented Kalman filtering.

[0031] The beneficial effects of this invention are as follows: By introducing frequency-hopping-based wireless ranging data between fixed anchor points and dual terminals (passenger and driver), and fusing multi-source data such as GNSS / IMU, a robust triangulation model is constructed, achieving high-precision relative positioning based on this model. Addressing the time delay and phase deviation inherent in frequency-hopping communication, this invention proposes using a sparse reconstruction algorithm to identify the stable path for each frequency band. Based on the stable path, time and phase calibration is performed for each frequency band. Utilizing the characteristic that multiplying the bidirectional channel frequency response of the subcarriers at the channel center frequency point can eliminate phase errors, this invention eliminates the global phase error and time deviation in frequency-modulated communication, obtaining accurate distance observations between the three nodes, thus providing a foundation for accurate prediction of the driver terminal's position.

[0032] Compared to traditional positioning methods that rely on a single GNSS signal, this invention significantly improves positioning accuracy and reliability. Especially in environments with weak satellite signals, such as urban canyons, it can still achieve stable relative positioning with an accuracy of less than 2 meters or even higher. This ensures that in application scenarios such as finding missing persons and ride-hailing pick-up, mobile drivers can quickly and accurately locate passengers, thereby improving the efficiency and safety of ride-hailing pick-up.

[0033] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0034] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0035] Figure 1This is a flowchart illustrating a relative positioning method based on channel splicing ranging in an urban canyon environment according to an embodiment of the present invention.

[0036] Figure 2 The simulation results of ranging accuracy after splicing different numbers of channels are shown in the figure.

[0037] Figure 3 The simulation results show the position estimation error between driver terminal C and passenger terminal B.

[0038] Figure 4 This is an example diagram illustrating an application scenario of the present invention. Detailed Implementation

[0039] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0040] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0041] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0042] One embodiment of the present invention provides a relative positioning method based on channel stitching ranging in an urban canyon environment. This method employs a three-node cooperative positioning structure, such as... Figure 4As shown, the system includes a fixed anchor point A, a passenger terminal B, and a driver terminal C. The fixed anchor point A is located at a fixed position with known coordinates (e.g., a building's podium, a ground signal tower, etc.). The passenger terminal B is carried by the passenger and remains relatively stationary (within a 1-2 meter range) during the positioning process. The driver terminal C is located on the vehicle and can obtain its own position via GNSS / IMU. This method uses frequency-hopping communication to stitch together multiple channels to achieve precise wireless ranging, acquiring distance data between the fixed anchor point A and passenger terminal B, between passenger terminal B and driver terminal C, and between the fixed anchor point A and driver terminal C. Combined with the positional information of the driver terminal C, nonlinear state estimation methods such as Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF) are used to accurately estimate the relative position of passenger terminal B.

[0043] like Figure 1 As shown, the method includes steps such as system initialization, ranging data acquisition, state prediction, observation update, and arrival judgment. It can continuously and in real time correct the position estimation of passenger terminal B, and finally achieve high-precision positioning of driver terminal C relative to passenger terminal B.

[0044] The specific details of this method are as follows:

[0045] 1. Initialization settings:

[0046] At the start of positioning, the initial position and velocity of driver terminal C (provided by GNSS / IMU) and the initial position estimate of passenger terminal B (e.g., obtained by satellite positioning of passenger mobile terminal, with a large error of 10 meters) are acquired, and the position of fixed anchor point A is known.

[0047] The position and speed of driver terminal C and the position of passenger terminal B are used as state variables to be estimated, and the filter is initially set. During initialization, it is assumed that passenger terminal B is stationary, and its initial position is set with a large uncertainty to cover possible actual position deviations.

[0048] Let the system state vector at time k be:

[0049]

[0050] in, These are the x and y coordinates of driver terminal C at time k, respectively. Let x and y be the velocity components of driver terminal C at time k; Let x and y be the coordinates of passenger terminal B at time k (assuming passenger terminal B is stationary and remains unchanged during prediction).

[0051] 2. Distance measurement data acquisition:

[0052] Points A, B, and C communicate via frequency hopping using OFDM wireless signals. Errors in the frequency hopping channel are eliminated, and multiple channels are stitched together to obtain channel state information from multiple aggregated frequency hopping channels without phase error. Further, compressed sensing algorithms, subspace decomposition, or maximum likelihood methods are used to estimate the signal's time of flight and perform ranging, acquiring distance measurements between each pair of nodes: A–B, B–C, and A–C. This distance data is periodically collected at a predetermined frequency (e.g., once per second).

[0053] Specifically:

[0054] 1) Frequency hopping channel data acquisition.

[0055] By performing rapid frequency hopping within a predetermined frequency band, channel state information (CSI) for multiple frequency bands is sequentially acquired in different time slots. Probe signals are transmitted sequentially on W frequency bands, and the received channel frequency response y(m,w) is measured, where each band has a bandwidth MΔf (containing M equally spaced subcarriers). This frequency hopping measurement effectively yields a frequency domain sampling channel with a total bandwidth of W×MΔf. Due to the influence of RF synchronization errors on the measurements of different frequency bands, unavoidable phase distortion is superimposed on each measured channel frequency response. These phase errors consist of two parts: one part varies linearly with the subcarrier index, exhibiting a "phase ramp" where the phase offset accumulates larger at higher frequencies; the other part is a constant phase offset independent of the subcarrier. These two types of deviations are mainly caused by sampling frequency deviation and time synchronization error (causing a linearly frequency-varying offset δ). w ) as well as carrier frequency offset and initial phase deviation (causing global constant offset σ) w This is caused by [the factor], and the phase deviation between these two parts is different between each frequency hopping channel. The measured channel frequency response model can be expressed as:

[0056]

[0057] Among them, f w,0 The reference frequency for frequency band w is... Let mΔf(τ) be the complex gain of the k-th path in frequency band w (including the amplitude and phase of the path). k +δ w ) represents the path delay τ k and the time offset δ of this frequency band w The phase term, f, which is caused by the linear variation with the subcarrier index, is w,0 τ k +σ wThis represents a constant phase term independent of the subcarrier (including the phase of the path at the reference frequency and the global phase offset of the band). K is the total number of paths, Δf is the subcarrier spacing, and n(m,w) is the Gaussian white noise on the m-th subcarrier of band w, where m = 1, 2, ..., M and w = 1, 2, ..., W.

[0058] Equation (2) shows that the CSI of different frequency bands have their own linear slope deviation and constant deviation in phase. If the multi-band data are directly spliced ​​together without correction, a broadband channel representation with consistent phase cannot be obtained. It should be noted that for frequency band w, a probe signal is transmitted on the frequency band to measure the channel frequency response (CFR) of a certain subcarrier, and the channel frequency responses of all subcarriers are combined to form the channel state information (CSI) of frequency band w.

[0059] 2) Stable path extraction (phase error identification).

[0060] To eliminate phase inconsistencies between CSI values ​​across different frequency bands, a stable path is first identified as a reference within the CSI of each band. A stable path refers to the same physical propagation path that exists and exhibits consistent characteristics across all frequency bands (typically the line-of-sight path (LoS) with the highest energy in each band). Taking advantage of the fact that the sparse components of multipath signals are far fewer than the number of frequency domain sampling points, a sparse reconstruction algorithm is preferred to estimate the principal path delay for each frequency band. For example, the CSI measurement y(·,w) of a single frequency band w can be represented as a linear combination of multipath parameters, and algorithms such as Orthogonal Matching Pursuit (OMP) are used to find the best-matching path in an overcomplete delay dictionary. The path with the highest energy is extracted through finite iterations, which is the stable path for that frequency band, and its estimated delay is denoted as . Initial phase is Since the number of dominant paths K is much smaller than the number of sampling points M′ (M′=M) in an indoor environment, the above sparse method can effectively extract the dominant stable path in each channel. The extracted stable path parameters provide a basis for subsequent correction of the phase error of each channel.

[0061] 3) Single-channel correction (eliminating phase errors between frequency bands).

[0062] Using the stable path of a specific frequency band (e.g., frequency band w0) identified in step 2) as a reference, time and phase calibration is performed on the CSI of other frequency bands. Specifically, for any frequency band w to be calibrated, the time delay difference between its stable path and the stable path of the reference frequency band w0 is calculated. Let w0 be the time delay of the stable path. δ is the time delay of the stable path in frequency band w. w This refers to the phase shift in frequency band w caused by sampling frequency deviation and time synchronization error. This refers to the phase shift in frequency band w0 caused by sampling frequency deviation and time synchronization error.

[0063] This Δt reflects the overall time offset of the CSI of frequency band w relative to the reference frequency band w0 (including the difference in delay between the two channel measurements and the difference in synchronization time deviation). Then, the entire CSI of frequency band w is equivalently time-shifted in the frequency domain by multiplying the spectrum of all its subcarriers by a linear phase factor exp(+j2πfΔt), thereby advancing the stable path delay of frequency band w by Δt to align it with the reference frequency band, where f is the subcarrier frequency. Simultaneously, a global phase rotation factor is applied to the CSI of frequency band w. Used to eliminate the initial phase difference of the stable path in this frequency band relative to the reference.

[0064] After the aforementioned time offset and phase rotation corrections, the stable path of frequency band w is aligned to the zero-delay point in the time domain, and its complex gain phase is zeroed. After performing the same calibration operation on all frequency bands, the stable paths of each channel are aligned to a common reference, eliminating absolute phase deviations between different frequency bands while maintaining the relative time delay relationships within each band. Let the CSI of the corrected frequency band w be denoted as... At this point, the phase references of CSI in each frequency band are consistent, laying the foundation for subsequent multi-channel splicing.

[0065] 4) Multi-channel splicing and relative channel reconstruction.

[0066] The corrected CSI data for each frequency band from step 3) are then stitched together and integrated to construct an equivalent broadband channel frequency response. Specifically, a frequency vector f = [f_band, f_subcarrier, f ... (1) ,f (2) ,...,f (WM) (M frequency bands, each with W subcarriers, therefore the length of the frequency vector f is WM), f (WM) Let M be the frequency vector of the Mth subcarrier in the Wth frequency band, and let CSI vector be the corresponding spliced ​​spectrum. in This represents a vector of M corrected CSI values ​​over frequency band W. The CSI obtained from the concatenated frequencies can be represented as a superposition of the contributions from each physical path:

[0067]

[0068] Among them, f m,w t represents the frequency of the m-th subcarrier in frequency band w. k ′ represents the equivalent delay after calibration for the k-th path, and c kLet this be the complex gain after path alignment. Clearly, the above equation is formally similar to the frequency domain sampling representation of the Discrete Fourier Transform (DFT). When all spliced ​​frequency points are equally spaced across the overall frequency band, for... The time-domain impulse response can be obtained by directly performing an IFFT.

[0069] However, because the phase alignment operation in step 3) applies an additional phase to each spectrum segment, the spliced ​​W-segment CSI is not a spectrum sampled at a strictly uniform interval. Therefore, the correct time delay component cannot be obtained directly using Fast Fourier Transform. To address this, the spliced ​​channel is considered as consisting of several frequency band blocks with unknown phase shifts, and the relative impulse response of the spliced ​​channel is reconstructed using a sparse inversion method. Specifically, the spliced ​​CSI vector is modeled as follows:

[0070]

[0071] Where F is the Fourier coefficient matrix constructed based on the frequency vector f (the size of the matrix is ​​determined by the number of candidate paths and the size of the frequency vector, and the k-th column in the matrix corresponds to the exp(-j2πft) generated by the delay of a certain candidate path). k ) sequence, t k Let r be the time delay of the k-th path or the flight time of the signal, r be the sparse impulse response coefficient vector to be determined, and n be the noise term. Then, the sparse solution of r can be obtained by solving the following optimization problem:

[0072]

[0073] Where β is the regularization parameter controlling sparsity, and ||·||² and ||·||₀ represent the L2 and L0 norms, respectively. The above optimization is equivalent to solving an inverse discrete Fourier transform, finding a few points in the continuous time-delay space where the frequency domain response matches the measurement. An iterative sparse reconstruction algorithm can be used to efficiently solve this problem.

[0074] In this embodiment, the "Soft Threshold Iterative Descent (STID)" method is preferably used to reduce computational complexity. The iterative process is as follows:

[0075] From the initial r (0) =0 to start iterative execution:

[0076] (a) Calculate the current residual gradient with F Where * denotes the conjugate transpose of the matrix, r (t) Let be the sparse impulse response coefficient vector for the t-th iteration;

[0077] (b) Determine the step size ε through line search;

[0078] (c) Update the estimate r (t+1) =r (t)+εg;

[0079] (d) for r (t+1) Perform soft thresholding to set coefficients smaller than the threshold βε to zero in order to maintain the sparsity of the results.

[0080] This process is iterated until the relative error or residual converges to a predetermined threshold. Through this sparse iterative reconstruction, the relative impulse response of the spliced ​​channel can be obtained. (The strongest stable path corresponds to a latency that has been aligned to 0). Compared to the bandwidth of a single channel, the bandwidth obtained by splicing multiple channels... With W times the bandwidth, it achieves W times the delay resolution improvement, and the position of each multipath in the impulse response can be more clearly distinguished.

[0081] 5) The final calibration yields an error-free CSI.

[0082] Step 4) obtained and its corresponding frequency domain response It is still a relative CSI (i.e., the stable paths in each frequency band are aligned but the actual absolute phase has not yet been recovered, and the propagation delay of the real signal path cannot be estimated). To obtain error-free CSI of the true physical channel, bidirectional channel measurement information is used to... Perform calibration.

[0083] For fixed anchor point A, passenger terminal B, and driver terminal C, bidirectional CSI is obtained between each pair of them. Specifically, based on the OFDM signal communication system, the transmitting and receiving nodes obtain the bidirectional CSI between each pair of fixed anchor point A, passenger terminal B, and driver terminal C through a fast bidirectional communication mechanism, i.e., the uplink (A→B) and downlink (B→A) CSI (A and C, B and C are the same). Since the uplink and downlink traverse the same physical channel, the multiplication of their complex frequency responses can cancel out the phase errors introduced by frequency offset, etc., to obtain the squared CSI, in which the phase error of the zero subcarrier (i.e., the center frequency of the channel) has been eliminated.

[0084] Therefore, the squared CSI value of the center frequency subcarrier in each frequency band of the communication links between A and B, A and C, and B and C is selected as the "real" channel frequency response sample for calibration. Let h be the actual CFR corresponding to these samples. 0,w The relative (uncalibrated) CFR is For any such calibration subcarrier, there are two main unknown phase offset components between its actual CFR and relative CFR: one is the frequency-dependent phase offset υ caused by the time reference offset (this offset υ corresponds to the equivalent time deviation at angular frequency 2πf), and the other is the frequency-independent global phase offset. (Overall phase difference of complex gain). The relationship is shown in the following equation:

[0085]

[0086] Where, υ=2πf 0,w t b , t b This represents the offset of the relative impulse response from the true impulse response on the time axis (i.e., the difference between the absolute arrival delay of the stable path in the actual channel and the delay relative to 0). As can be seen from equation (6), the parameter... A fixed phase shift affects all frequency points, while the parameter υ varies linearly with frequency.

[0087] Using the actual and relative CFR values ​​at the selected center frequency subcarrier (m=0) in each frequency band, a system of equations is established to solve for the global phase correction parameters. and time deviation parameter t b Specifically, the reference values ​​obtained by squaring the CSI values ​​of each frequency band are compared with the corresponding values ​​calculated by relative CSI, and an error function is constructed. For example, accumulate the sum of squared phase differences of the center subcarriers in each frequency band, and then find the value that minimizes the error function. and t b .

[0088] This optimization problem is a nonlinear least squares problem, which can be solved efficiently using numerical algorithms such as the conjugate gradient method to obtain an estimate. and Finally, the calibration factor is applied to the entire spliced ​​spectrum. Above, for each subcarrier frequency point f m,w Multiply by the corresponding correction phase The actual CSI spectrum H after removing synchronization errors can then be obtained. The corresponding relationship is shown in the following formula:

[0089]

[0090] Among them, H m,w This represents the true channel frequency response of the m-th subcarrier in the calibrated frequency band w. Thus, after the above steps, a stitched CSI spectrum H, free from the effects of frequency hopping channel errors, is obtained. The phase of each subcarrier in this spectrum is consistent with the actual physical channel, and it does not contain phase shifts caused by frequency switching or synchronization errors. The error-free stitched CSI can be used for subsequent accurate extraction of multipath delay and channel parameters, and for estimating the signal's time of flight using compressed sensing algorithms, subspace decomposition, or maximum likelihood methods. This time-of-flight estimation is then multiplied by the speed of light to convert it into the precise distance between the transceiver ends.

[0091] 3. State prediction:

[0092] The state of driver terminal C at the next moment is predicted based on its motion model. It is generally assumed that driver terminal C moves at a constant linear velocity within two ranging cycles, based on the velocity from the previous moment. Therefore, the current position of driver terminal C can be predicted by adding the estimated position from the previous moment to the velocity and multiplying by the time interval. Passenger terminal B, being stationary, maintains the same position as its previous estimate during the prediction phase. Thus, the predicted state value x for the current moment is obtained. k∣k-1 This includes the predicted location and speed of driver terminal C, and the predicted location of passenger terminal B.

[0093] Specifically, using a discrete linear motion model, assuming a sampling interval Δs = 1s, then:

[0094] x k∣k-1 =Gx k-1∣k-1 +w k-1 ,

[0095] Among them, w k-1 For process noise, x k-1∣k-1 To update the system state x after the observation is completed at time k-1 k-1 The posterior estimate is given by Q, where Q is the covariance matrix of the process noise. G is the state transition matrix.

[0096]

[0097] 4. Observation Update:

[0098] The predicted state is fused and corrected with real-time ranging observations. Based on the ranging data obtained in step 2, the actual observed distances A–B, B–C, and A–C are calculated and compared with the theoretical distances calculated based on the predicted state values ​​to obtain the observation residuals. Then, nonlinear filtering algorithms such as Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF) are used to correct the predicted state based on the above residuals. After observation updates, the corrected position estimates for driver terminal C and passenger terminal B can be obtained. The filtering process can reduce the positioning error caused by measurement errors and environmental interference, making the position estimate of passenger terminal B gradually approach the true position.

[0099] The update process includes:

[0100] 1) Observation vector: At time k, it is obtained through the channel splicing ranging technique in step 2:

[0101]

[0102] Among them, z k d is the actual measurement and observation vector at the current time k; AC,k d is the distance between fixed anchor point A and driver terminal C.BC,k d represents the distance between passenger terminal B and driver terminal C. AB,k The distance between fixed anchor point A and passenger terminal B.

[0103] 2) Observation equation: The actual observed value is determined by the state and is superimposed with measurement noise v. k :

[0104] z k =l(x k )+v k ,

[0105] Where R is the observation noise covariance matrix. l(·) consists of the following three distance functions:

[0106]

[0107] x A y A The x and y coordinates of the fixed anchor point A.

[0108] 3) Filtering Recursion: The above nonlinear system is recursively refuted using either Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF), mainly including:

[0109] predict:

[0110]

[0111] in, For the state estimation at time k-1, Let P be the prior state at the current time k. k-1∣k-1 Let P be the posterior error covariance matrix. k∣k-1 Let G be the prior error covariance matrix. In state prediction, the state transition matrix G is used to estimate the state at the previous time step. Extrapolate to obtain the prior state at the current moment. Simultaneously, the estimation error covariance P is propagated to calculate the prior covariance P. k∣k-1 To quantify the uncertainty of the prediction, a process noise covariance matrix Q is introduced into the prediction to model external disturbances and modeling errors in the dynamic process of the system.

[0112] renew:

[0113] Update observation residuals:

[0114] Update the observation matrix: UKF is generated from sampling points.

[0115] Update Kalman gain: K k =P k|k-1L k T (L k P k|k-1 L k T +R) -1

[0116] Update post-verification status:

[0117] Updated post-hoc covariance: P k|k =(IK k L k )P k|k-1 I is the identity matrix

[0118] The EKF or UKF algorithm steps (Jacobian calculation, sigma point generation, etc.) involved in the above formulas are all implemented using existing technologies, and will not be described in detail in this embodiment.

[0119] 5. Repeated iterations:

[0120] The process described in steps 2 to 4 is continuously executed to update the position estimate of passenger terminal B in real time. As driver terminal C moves and acquires new ranging data, state prediction and observation updates are iterated repeatedly. After each iteration, the position accuracy of passenger terminal B will improve, thereby achieving continuous tracking of the passenger terminal's position.

[0121] 6. Arrival Judgment:

[0122] After each location update, it is determined whether driver terminal C has reached the vicinity of passenger terminal B. When the distance between driver terminal C and passenger terminal B, estimated by filtering, is less than a preset threshold (e.g., 20 meters) and remains so for several measurement cycles, it can be determined that the driver has approached and reached the passenger's location. At this point, driver terminal C can be notified to end navigation, and passenger terminal B can be prompted to board the vehicle. The positioning process is then complete.

[0123] Specifically, the distance between driver terminal C and passenger terminal B obtained by filtering estimation is expressed as:

[0124]

[0125] in, These are the x and y coordinates of the driver terminal C obtained through filtering estimation, respectively. These are the x and y coordinates of passenger terminal B obtained through filtering estimation. If... If the distance is less than a preset threshold (e.g., 20m) and continues for more than N measurement cycles, it is determined that the driver has reached the passenger's location, triggering a prompt to end the process.

[0126] Simulations were performed on the method described in this embodiment, and the results are as follows: Figure 2 and Figure 3 As shown.

[0127] Figure 2 The relationship between ranging accuracy and the number of stitched channels is illustrated. It can be seen that the ranging accuracy improves with the increase in the number of stitched channels. Without stitching channels (i.e., one channel), the error in the CDF diagram of ranging error has a 90% probability of 2.14m. As the number of stitched channels increases, the ranging error decreases significantly: with two stitched channels, the 90% probability error is 0.82m; with three stitched channels, the 90% probability error is 0.32m; and with four stitched channels, the 90% probability error is only 0.15m.

[0128] Figure 3 The diagram shows the change in positional error between passenger terminal B and driver terminal C over time. It can be seen that as time increases and the driver gradually approaches the passenger, the errors between B and C and their actual positions gradually decrease, with both errors remaining around 2 meters.

[0129] In summary, this invention proposes a relative positioning method based on channel stitching ranging, which can achieve precise relative positioning between communication nodes. It enables high-precision relative positioning between ride-hailing drivers and passengers with an error of less than 2 meters, even in complex environments such as urban canyons. This invention fully utilizes anchor point reference information and wireless ranging data, effectively compensating for the shortcomings of satellite navigation in urban environments and providing reliable positioning support for applications such as passenger boarding and finding people.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A relative positioning method based on channel stitching ranging in an urban canyon environment, characterized in that, A three-node collaborative positioning structure is adopted, with fixed anchor points, passenger terminals and driver terminals set up to obtain the current position and speed information of the nodes. Frequency hopping communication is performed using OFDM wireless signals between fixed anchor points, passenger terminals, and driver terminals to collect channel status information of multiple frequency bands; A sparse reconstruction algorithm is used to identify a stable path in each frequency band. The stable path of one frequency band is selected as a reference, and the channel state information of other frequency bands is calibrated in terms of time and phase to initially eliminate frequency hopping channel errors. The channel state information of each frequency band after correction is spliced ​​and integrated to construct an equivalent broadband channel frequency response; then the spliced ​​channel is regarded as composed of several frequency band blocks with unknown phase shift, and the relative impulse response of the spliced ​​channel is reconstructed through sparse inversion. Two-way channel state information is obtained through two-way communication between the fixed anchor point, passenger terminal, and driver terminal. The two-way channel frequency response of each channel center frequency subcarrier is multiplied to obtain the squared CSI value, which is used as the channel frequency response sample for each frequency band. Based on the actual channel frequency response value and relative channel frequency response value of each frequency band channel frequency response sample, an error function is constructed to solve for the error parameters. Then, the frequency hopping channel error is eliminated based on the error parameters to obtain error-free spliced ​​channel state information. The flight time of the signal is estimated based on error-free spliced ​​channel state information to obtain the distance measurements between each pair of the fixed anchor point, passenger terminal and driver terminal, wherein distance data is periodically collected at a predetermined frequency. The position of the driver terminal at the next moment is predicted using a discrete linear motion model. The measured distances between each pair of the fixed anchor point, passenger terminal, and driver terminal are compared with the theoretical distances between each pair of the fixed anchor point, passenger terminal, and driver terminal calculated based on the position prediction to obtain the observation residuals. Based on the observation residuals, Kalman filtering is used to correct the predicted position of the passenger terminal. Repeat the above process until the estimated distance between the driver terminal and the passenger terminal is less than a preset threshold, and maintain this for multiple measurement cycles, then determine that the driver terminal has reached the location of the passenger terminal.

2. The method according to claim 1, characterized in that, A sparse reconstruction algorithm is used to identify a stable path in each frequency band. For any frequency band, its channel frequency response is represented as a linear combination of multipath parameters. An orthogonal matching pursuit algorithm is used to find the best matching path in an overcomplete time delay dictionary. The path with the maximum energy is extracted through finite iterations, which is the stable path of that frequency band.

3. The method according to claim 2, characterized in that, After identifying a stable path for each frequency band, select one stable path of frequency band w0 as a reference benchmark, and calculate the time delay difference between the stable paths of other frequency bands w and the stable paths of frequency band w0. The estimated time delay for the stable path in frequency band w0. δ is the estimated time delay of the stable path in frequency band w. w This refers to the phase shift in frequency band w caused by sampling frequency deviation and time synchronization error. The phase shift in frequency band w0 caused by sampling frequency deviation and time synchronization error; Multiply the spectrum of all subcarriers in frequency band w by a linear phase factor exp(+j2πfΔt) to advance the stable path delay of frequency band w by Δt, thereby aligning it with the stable path of frequency band w0, where f is the frequency of the subcarrier; simultaneously, apply a global phase rotation factor to the CSI composed of the channel frequency response of all subcarriers in frequency band w. Eliminate the initial phase difference between the stable path of frequency band w and the stable path of frequency band w0. The initial phase of the stable path for frequency band w.

4. The method according to claim 3, characterized in that, The corrected channel state information of each frequency band is concatenated and integrated to construct an equivalent broadband channel frequency response, that is, to form a frequency vector f = [f] containing all frequency band subcarriers. (1) ,f (2) ,…,f (WM) ], f (WM) Let M be the frequency vector of the Mth subcarrier in the Wth frequency band, and the corresponding concatenated CSI vector. in This represents a vector of M corrected CSI values ​​over frequency band W; The spliced ​​channel is considered as consisting of several frequency band blocks with unknown phase shifts. The relative impulse response of the spliced ​​channel is reconstructed using a sparse inversion method, and the spliced ​​CSI vector is modeled as follows: In the formula, F is the Fourier coefficient matrix constructed based on the frequency vector f, r is the sparse impulse response coefficient vector to be determined, and n is the noise term; the sparse solution of r is obtained by solving the following optimization problem: In the formula, β is the regularization parameter that controls sparsity, and ||·|2 and ||·|0 represent the L2 and L0 norms, respectively. The soft threshold iterative descent method is used to solve this optimization problem.

5. The method according to claim 4, characterized in that, The soft threshold iterative descent method is used to solve the optimization problem. This includes: calculating the gradient between the residual and the Fourier coefficient matrix F at iteration t. * denotes the conjugate transpose of the matrix, r (t) Let r be the sparse impulse response coefficient vector of the t-th iteration; determine the step size ε through line search; update the estimated r. (t+1) =r (t) +εg;for r (t+1) A soft thresholding process is performed, setting coefficients smaller than the threshold βε to zero to maintain the sparsity of the results. Through iterative iteration, the process stops when the relative error or residual converges to a predetermined threshold, yielding the relative impulse response of the spliced ​​channel.

6. The method according to claim 5, characterized in that, relative impulse response Perform FFT to obtain the spectrum Then, the bidirectional channel frequency response is used to analyze the spectrum. Calibration includes: Bidirectional Channel Frequency Response (CSI) between the fixed anchor point, passenger terminal, and driver terminal is obtained through a fast two-way communication mechanism. The bidirectional channel frequency response of each channel center frequency subcarrier is multiplied to obtain the squared CSI value, which is then used as a channel frequency response sample for each frequency band. The actual channel frequency response value and the relative channel frequency response value of each frequency band sample are compared to construct an error function. Solving this error function yields the global phase offset estimate. and time deviation parameter estimates For each subcarrier frequency point f m,w Multiply by the corrected phase Obtain the calibrated true channel frequency response H m,w The actual CSI spectrum H, after removing frequency hopping channel errors, is obtained by combining the real channel frequency responses of each subcarrier, where m = 1, 2, ..., M, M is the number of subcarriers in each frequency band, and w = 1, 2, ..., W, W is the number of frequency bands.

7. The method according to claim 6, characterized in that, The flight time of the signal is estimated based on the spectrum H to obtain the distance measurements between each pair of the fixed anchor point, passenger terminal, and driver terminal; The driver terminal's position at the next moment is predicted using a discrete linear motion model, which is expressed as follows: where w k-1 is the process noise, G is the state transition matrix, x k-1∣k-1 is the a posteriori estimate of the system state x k-1 , and Q is the process noise covariance matrix. The measured distances between each pair of the fixed anchor point, passenger terminal, and driver terminal are compared with the theoretical distances between each pair of the fixed anchor point, passenger terminal, and driver terminal calculated based on the location prediction values ​​to obtain the observation residuals. Based on the observation residuals, extended Kalman filtering or unscented Kalman filtering is used to correct the position prediction values ​​of the passenger terminal.

Images