A millimeter wave channel tracking method based on reconfigurable smart surface
By building a dual-time scale model and angle change model in a millimeter wave channel, combining Fourier transform and maximum likelihood algorithm for channel angle tracking, the problem of high channel angle tracking in the existing technology is solved, and low complexity is achieved channel tracking improves communication reliability.
Patent Information
- Application Number
- CN202210805376.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-10
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2042-07-10
AI Technical Summary
The prior art is difficult to effectively track the channel angle parameters between the user's end and the smart surface in a millimeter wave channel, especially when the user moves slowly or the scatterer suddenly changes, resulting in high computational complexity and unable to achieve low complexity. Channel tracking.
Using a millimeter wave channel tracking method based on reconstructible intelligent surfaces, the channel angle parameters are initialized by building a dual-time scale model and angle change model, and the channel angle parameters are initialized by using a two-dimensional fast Fourier transform algorithm, and the maximum likelihood algorithm and Newtonian algorithm are combined to perform channel angles. Track and reinitialize parameters when a mutation in the channel matrix is detected to reduce computational complexity.
On the premise of ensuring superior performance, the computing complexity is significantly reduced, computing power resources are saved, and the reliability of RIS-assisted wireless communication is improved. It is suitable for channel tracking in scenarios such as slow movement of the user side or vehicles and pedestrians.
Smart Images

Figure CN115412187B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of mobile communication technology, and in particular to a millimeter wave channel tracking method based on a reconfigurable smart surface. Background Art
[0002] As one of the potential technologies for 6G mobile communications, Reconfigurable Intelligent Surface (RIS) has broad application prospects. RIS-assisted massive Multiple Input Multiple Output (MIMO) systems can significantly improve wireless communication coverage, throughput, and spectrum efficiency. However, the high loss characteristics of millimeter wave channels make it easy to block the line of sight (LoS) path between user equipment (UE) and base station (BS), which seriously affects the communication quality.
[0003] The introduction of RIS can dynamically adjust the signal amplitude or phase shift of each reflection unit to improve this problem, but it requires accurate channel state information (CSI) of the UE-RIS-BS link as support. Because the spatial positions of the BS and RIS remain unchanged, the millimeter wave channel on the RIS-BS side can be modeled as a quasi-static channel, while the slow movement of the user end or the sudden blockage of scatterers such as vehicles and pedestrians will cause the UE-RIS channel to suffer from channel mutations and slow angle changes. Therefore, tracking the UE-RIS channel angle parameters in multiple time slots to ensure estimation performance, and separately estimating the UE-RIS-BS cascade channel, reducing unnecessary channel estimation overhead to reduce computational complexity have become the current research focus.
[0004] At present, a large number of literatures have studied RIS-assisted millimeter wave channel estimation, but most of the existing studies do not consider the channel tracking problem when there are angle changes, and are not universal for the practical application of RIS. Some other studies have considered the changes of UE-RIS channels over time, but the algorithms used can only track the entire cascade channel and cannot separate and estimate a single channel, resulting in huge computational complexity. It can be seen that it is of great practical significance to study low-complexity channel tracking solutions in scenarios where scatterers change suddenly and users move slowly. Summary of the invention
[0005] In view of this, the purpose of the present invention is to provide a millimeter wave channel tracking method based on a reconfigurable smart surface, which can achieve lower computational complexity and greatly save computing resources while ensuring superior performance.
[0006] The present invention solves the above technical problems by the following technical means:
[0007] A millimeter wave channel tracking method based on a reconfigurable smart surface is applied to the millimeter wave channel tracking problem in an environment where scatterers suddenly change and the user end moves slowly, including the following steps:
[0008] S1. Construct a RIS-assisted uplink wireless communication system model, as well as a dual-time scale model and an angle variation model of the UE-RIS-BS link;
[0009] S2. Initialize the channel angle parameters using a two-dimensional fast Fourier transform algorithm and initialize the path gain parameters using a maximum likelihood algorithm;
[0010] S3, tracking the channel angle in the horizontal direction of UE-RIS based on Newton algorithm;
[0011] S4. Detect whether the channel matrix H(t) will mutate. If a mutation is detected in the channel matrix H(t), initialize the parameters again. Otherwise, use the Newton algorithm to continue tracking the angle parameters.
[0012] Furthermore, since the channel of the UE-RIS-BS link can only communicate through RIS end reflection, the method of step S1 specifically includes the following steps:
[0013] S101, equip BS with N R Uniform linear array antennas, serving N T For a single user with one antenna, the RIS is composed of Reflection units, among which, and are the number of reflection elements in the horizontal and vertical directions of RIS, respectively;
[0014] S102, L RF chains are connected horizontally and vertically in RIS, and the LoS path from UE to BS is blocked. The signal sent by UE in the tth time slot can only be reflected to the BS end through RIS, and the received signal at the BS end is for:
[0015] Y(t)=GH(t)Γ(t)+N(t)
[0016] in, is the channel matrix of RIS-BS, is the channel matrix of UE-RIS, is the pilot signal sent by the UE, T is the number of time slots, and the pilot signal meets the orthogonal condition The mean is 0 and the variance is Complex Gaussian white noise;
[0017] S103 , assuming that the millimeter wave channel G of RIS-BS is a quasi-static channel, model the channel matrices G and H(t) of UE-RIS and RIS-BS respectively.
[0018] Further, the channel matrices G and H(t) are modeled based on the dual time scale model. In step S103, the UE-RIS and RIS-BS channel matrices are modeled as follows:
[0019]
[0020] Among them, L a and L b are the number of paths between UE-RIS and RIS-BS, respectively. is the lth channel H(t) a The path gain, is the lth channel G b The path gain, is the UE Angle of Departure (AoD), and are the elevation and azimuth of the RIS angle of arrival (AoA), and are the elevation and azimuth of the AoD of RIS, is the AoA of BS, and is the array steering vector of channel H(t).
[0021] Furthermore, the angle change is modeled based on the Markov chain, and the array steering vector of the channel H(t) and They are:
[0022]
[0023] in, d U and d I are the antenna distances between UE and RIS, respectively, and λ is the wavelength of electromagnetic waves;
[0024] Expressions and Similar, and d U Use and d B Instead, d B is the BS antenna spacing, Expressions and Similar, and Use and replace;
[0025] The angle parameter set O(t) of all UE-RIS paths in the tth time slot is:
[0026]
[0027] Since the channel angle change between adjacent time slots can be modeled as a first-order Markov chain, the angle model when there is an angle change in the t-th time slot is:
[0028] O(t)=O(t-1)+u(t)
[0029] in, Indicates that the mean is 0 and the variance of the angle change is The complex Gaussian white noise, Q U is the covariance matrix of u(t).
[0030] Further, in step S2, a two-dimensional fast Fourier transform algorithm is used to initialize the channel angle parameters, including RIS estimating the angle in the horizontal direction and RIS estimating the angle in the vertical direction. The method of estimating the angle in the horizontal and vertical directions of the RIS is the same. The RIS estimating the angle in the horizontal direction specifically includes the following steps:
[0031] S201, RIS horizontal receiving signal for:
[0032]
[0033] in, is the horizontal channel matrix from UE to RIS, N x (t) is Gaussian white noise, and the received signal is decorrelated with the pilot signal and then filled with zero to obtain
[0034] S202, obtaining the signal spatial spectrum through 2D-FFT Specifically:
[0035]
[0036] in, H x The spatial spectrum of (t), To obtain the Gaussian white noise N′ superimposed on the received signal after zero padding x The spatial spectrum of (t), F L′ and is the Fourier matrix after zero padding, the channel matrix The element in row m and column n of for:
[0037]
[0038] Among them, the definition After peak search, the l a The coordinates of the peak point of the path are get Therefore, the first horizontal direction from UE to RIS a The angle estimate for this path is:
[0039]
[0040] S203, perform 2D-FFT on the vertical direction received signal of the RF chain connected from the UE to the RIS to estimate for:
[0041]
[0042] in, for The corresponding ordinate of the peak point.
[0043] Furthermore, in step S2, the path gain parameter is initialized using the maximum likelihood algorithm, specifically:
[0044] Rewrite the signal model as:
[0045]
[0046] in, H x (t) Array steering matrix and the path gain matrix They are:
[0047]
[0048] Using ML algorithm to solve the estimated value of the path gain parameter of the UE-RIS channel matrix in the tth time slot for:
[0049]
[0050] in,(·) -1 To invert the matrix, and is the estimated value of the array steering matrix. Similarly, the RIS-BS channel path gain can be initialized.
[0051] Further, the method of step S3 specifically includes the following steps:
[0052] S301, after tracking and obtaining the angle parameter of a path, remove the received signal of the path to obtain the residual F x (t) is:
[0053]
[0054] in, and For the front a The estimated channel matrix parameters obtained by the iteration simplifies the residual energy R(t) to:
[0055]
[0056] Among them, l a =1,...,L a is the number of iterations, and The first a After iterations and estimated values The column vector of minimized residual energy R(t) is obtained and The value of can be used to trace the channel angle parameter value;
[0057] S302, R(t) The Taylor expansion is:
[0058]
[0059] Among them, R u ′(t) and R v ′(t) is the first partial derivative of function R(t) with respect to u(t) and v(t), R u ″ u (t),R u ″ v (t),R v ″ u (t) and R v ″ v (t) are the second partial derivative of function R(t) with respect to u(t), the partial derivative of u(t) followed by the partial derivative of v(t), the partial derivative of v(t) followed by the partial derivative of u(t), and the second partial derivative of v(t);
[0060] S303, transform the UE-RIS channel angle parameter tracking problem into solving R u ′(u(t),v(t))=0 and R v The problem of (u(t), v(t)) value when ′(u(t), v(t)) = 0, R u ′(u(t),v(t)) and R v ′(u(t),v(t)) is:
[0061]
[0062] S304, solve R by combining u ′(u(t),v(t))=0,R v ′(u(t),v(t))=0, we get:
[0063]
[0064] Because in the l a The channel angle parameter value obtained by looping only once in a tracking operation may not be the most accurate, so multiple loops are required. Assuming that loop R c Then the first a The channel angle parameter value obtained in the kth cycle of tracking is:
[0065]
[0066] Where k = 1,…,R c , is the first-order partial derivative, is the second-order partial derivative;
[0067] S305 , iteratively tracking the channel angle parameter value, the tracking value convergence condition is that the residual is less than the required accuracy value, and after the angle parameter value is tracked, the path gain parameter value is estimated using ML.
[0068] S306: Use the channel angle parameter value and the path gain parameter value to restore the channel matrix H of the UE-RIS horizontal RF chain in the tth time slot x (t).
[0069] Similarly, the channel matrix H of the RF chain in the vertical direction of UE-RIS is y The (t) parameter can also be estimated by tracking using the above method.
[0070] Further, the method of step S4 specifically includes the following steps:
[0071] S401. Define the likelihood function as:
[0072]
[0073] Among them, Q U is the noise covariance matrix, To calculate the matrix Q U Determinant N T When the H0 event occurs, Approaching 0;
[0074] S402. According to the Newman-Pearson criterion, the log-likelihood ratio is:
[0075]
[0076] When the judgment is an H1 event, the following conditions must be met: Where Υ is the predefined decision threshold, and the false alarm probability P of an H0 event being judged as an H1 event is preset. fa for:
[0077]
[0078] Given the false alarm probability P fa The value of the threshold Υ can be determined. Chi-square distribution definition is the right tail probability, where p(t) is the probability density function, so the calculation threshold Υ is:
[0079]
[0080] S403, according to the decision criteria in step S402, when it is determined to be an H1 event, it can be considered that the channel has a sudden change, at this time the channel matrix is reinitialized, and then the next tracking process begins.
[0081] It should be noted that in the present invention, RIS represents reconfigurable intelligent surface, UE represents user, BS represents base station, LoS represents line of sight, CSI represents channel state information, RF represents radio frequency, and 2D-FFT represents two-dimensional fast Fourier transform.
[0082] Beneficial effects of the present invention:
[0083] 1. The present invention connects the components of RIS to the RF chain, first uses the two-dimensional fast Fourier transform algorithm to initialize the estimated angle, and then uses the maximum likelihood algorithm to estimate the path gain; in the subsequent time slots, the Newton algorithm is used to track the angle parameters of each time slot. Since the high path loss of the millimeter wave channel and the sudden change of the scatterer may cause the channel matrix to mutate, if a mutation is detected, the parameters are initialized again, otherwise the Newton algorithm is used to continue tracking the angle parameters. Compared with the existing solutions, the channel tracking solution based on the Newton algorithm of the present invention can achieve lower computational complexity while ensuring superior performance, greatly saving computing resources.
[0084] 2. The present invention also considers the change of UE-RIS link angle and channel mutation caused by slow movement of the user end or sudden blockage of scatterers such as vehicles and pedestrians, which is more practical.
[0085] 3. The present invention utilizes a channel mutation detection scheme to detect UE-RIS channels in different time slots to improve the reliability of RIS-assisted wireless communication. BRIEF DESCRIPTION OF THE DRAWINGS
[0086] Figure 1 It is a flow chart of a millimeter wave channel tracking method based on a reconfigurable smart surface of the present invention;
[0087] Figure 2 It is a schematic diagram of the RIS assisted uplink wireless communication system;
[0088] Figure 3 It is a comparison curve of the computational complexity and the number of user antennas of the present invention, LS, LMMSE and KF algorithms;
[0089] Figure 4 This is the NMSE and SNR comparison curve of the present invention and different algorithms such as LS, LMMSE and KF;
[0090] Figure 5 This is a comparison curve of NMSE and time slot of different algorithms including LS, LMMSE and KF. DETAILED DESCRIPTION
[0091] The present invention will be described in detail below with reference to the accompanying drawings:
[0092] like Figure 1 As shown, a millimeter wave channel tracking method based on a reconfigurable smart surface in this embodiment specifically includes the following steps:
[0093] S1. Construct a RIS-assisted uplink wireless communication system model, as well as a dual-time scale model and an angle variation model of the UE-RIS-BS link.
[0094] like Figure 2 As shown, the BS is equipped with N R Uniform Linear Array (ULA) antennas, serving N T A single user with 1 antenna. RIS consists of Reflection units, and are the number of reflection elements in the horizontal and vertical directions of the RIS, respectively. Since the RIS does not have signal processing capabilities, this embodiment considers connecting L RF chains in the horizontal and vertical directions of the RIS, and the LoS path from the UE to the BS is blocked. The signal sent by the UE in the tth time slot is reflected by the RIS to the received signal at the BS end. for:
[0095] Y(t)=GH(t)Γ(t)+N(t)
[0096] in, is the channel matrix of RIS-BS, is the channel matrix of UE-RIS, is the pilot signal sent by the UE, T is the number of time slots, and the pilot signal meets the orthogonal condition The mean is 0 and the variance is Complex Gaussian white noise.
[0097] In this system, since the spatial positions of BS and RIS remain unchanged, it is assumed that the millimeter wave channel G of RIS-BS is a quasi-static channel, and the UE-RIS link may cause the millimeter wave channel H(t) to change suddenly and the angle to change slowly due to the slow movement of the user end or the sudden blockage of scatterers such as vehicles and pedestrians. Therefore, the channel matrices of UE-RIS and RIS-BS are modeled as:
[0098]
[0099] Among them, L a and L b are the number of paths between UE-RIS and RIS-BS, respectively. is the lth channel H(t) a The path gain, is the lth channel G b The path gain, is the UE departure angle (Angle of Departure, AoD), and are the elevation and azimuth of the RIS angle of arrival (Angle of Arrival, AoA), and are the elevation and azimuth of the AoD of RIS, is the AoA of the BS, the array steering vector of the channel H(t) and They are:
[0100]
[0101] in, d U and d I are the antenna distances between UE and RIS, and λ is the wavelength of electromagnetic waves. An expression similar to Will and d U Use and d B Instead, d B is the BS antenna spacing, An expression similar to Will and Use and Instead, the angle parameter set O(t) of all UE-RIS paths in the tth time slot is:
[0102]
[0103] Since the channel angle change between adjacent time slots can be modeled as a first-order Markov chain, the angle model when there is an angle change in the t-th time slot is:
[0104] O(t)=O(t-1)+u(t)
[0105] in, Indicates that the mean is 0 and the variance of the angle change is Complex Gaussian white noise.
[0106] S2. Initialize the channel angle parameters based on the 2D-FFT algorithm, and initialize the path gain parameters based on the maximum likelihood algorithm.
[0107] Because the method of estimating angles in the horizontal and vertical directions of RIS is the same, the present invention takes the horizontal receiving signal of the RF chain connected to RIS as an example to illustrate the initialization of angle parameters. for:
[0108]
[0109] in, is the horizontal channel matrix from UE to RIS, N x (t) is Gaussian white noise, and the received signal is decorrelated with the pilot signal and then filled with zero to obtain Then use 2D-FFT to get the signal spatial spectrum for:
[0110]
[0111] in, H x The spatial spectrum of (t), To obtain the Gaussian white noise N′ superimposed on the received signal after zero padding x (t) spatial spectrum. L′ and is the Fourier matrix after zero padding, the channel matrix The element in row m and column n of for:
[0112]
[0113] Among them, the definition After peak search, the l a The coordinates of the peak point of the path are get Therefore, the first horizontal direction from UE to RIS a The angle estimate for this path is:
[0114]
[0115] Performing 2D-FFT on the vertical received signal of the RF chain from UE to RIS can estimate for:
[0116]
[0117] in, for The corresponding ordinate of the peak point can be used to initialize the angle parameter of RIS-BE channel G in the same way.
[0118] Using the ML algorithm to initialize the path gain parameters, the signal model can be rewritten as:
[0119]
[0120] in, H x (t) Array steering matrix and the path gain matrix They are:
[0121]
[0122] Using ML algorithm to solve the estimated value of the path gain parameter of the UE-RIS channel matrix in the tth time slot for:
[0123]
[0124] in,(·) -1 To invert the matrix, and is the estimated value of the array steering matrix. Similarly, the RIS-BS channel path gain can be initialized.
[0125] S3. Track the channel angle in the horizontal direction of UE-RIS based on the Newton algorithm. After tracking and obtaining the angle parameter of a path, remove the received signal of the path to obtain the residual error:
[0126]
[0127] in, and For the front a The estimated channel matrix parameters obtained by the iteration simplifies the residual energy R(t) to:
[0128]
[0129] Among them, l a=1,...,L a is the number of iterations, and The first a After iterations and estimated values Minimizing the residual energy R(t) yields and The value of can be traced to the channel angle parameter value, and R(t) is The Taylor expansion is:
[0130]
[0131] Among them, R u ′(t) and R v ′(t) is the first partial derivative of function R(t) with respect to u(t) and v(t), R u ″ u (t),R u ″ v (t),R v ″ u (t) and R v ″ v (t) are the second partial derivative of the function R(t) with respect to u(t), the partial derivative of u(t) followed by the partial derivative of v(t), the partial derivative of v(t) followed by the partial derivative of u(t), and the second partial derivative of v(t).
[0132] In this embodiment, the UE-RIS channel angle parameter tracking problem is transformed into solving R u ′(u(t),v(t))=0 and R v The problem of (u(t), v(t)) value when ′(u(t), v(t)) = 0, R u ′(u(t),v(t)) and R v ′(u(t),v(t)) is:
[0133]
[0134] By solving R u ′(u(t),v(t))=0,R v ′(u(t),v(t))=0, we get:
[0135]
[0136] Because in the l a The channel angle parameter value obtained by looping only once in a tracking operation may not be the most accurate, so multiple loops are required. Assuming that loop R c Then the first aThe channel angle parameter value obtained in the kth cycle of tracking is:
[0137]
[0138] Where k = 1,…,R c , is the first-order partial derivative, is the second-order partial derivative.
[0139] The above formula is used to iteratively track the channel angle parameter value. The tracking value convergence condition is that the residual is less than the required accuracy value. After the angle parameter value is tracked, the path gain parameter value is estimated using ML. Finally, the channel angle parameter value and the path gain parameter value are used to restore the channel matrix H of the RF chain in the horizontal direction of the UE-RIS in the tth time slot. x (t). Similarly, the channel matrix H of the RF chain in the vertical direction of UE-RIS y The (t) parameter can also be estimated by tracking using the above method.
[0140] S4. Check whether the channel matrix H(t) will mutate. First, there are two hypotheses: H1: the channel matrix H(t) mutates; H0: the channel matrix H(t) does not mutate. The likelihood function is defined as:
[0141]
[0142] Among them, Q U is the noise covariance matrix, To calculate the matrix Q U Determinant N T When the H0 event occurs, Approaches 0. According to the Newman-Pearson criterion, the log-likelihood ratio is:
[0143]
[0144] When the judgment is an H1 event, the following conditions must be met: Where Υ is the predefined decision threshold, and the false alarm probability P of an H0 event being judged as an H1 event is preset. fa for:
[0145]
[0146] Given the false alarm probability P fa The value of the threshold Υ can be determined. Chi-square distribution definition is the right tail probability, where p(t) is the probability density function, so the calculation threshold Υ is:
[0147]
[0148] According to the above judgment criteria, when the judgment is H1 event, it can be considered that the channel has a sudden change. At this time, the channel matrix is reinitialized and then the next tracking process begins.
[0149] The simulation is performed according to the method of the present invention and compared with the LS, LMMSE and KF algorithms. The results are as follows: Figure 3 , Figure 4 and Figure 5 As shown:
[0150] from Figure 3 It can be seen that the computational complexity of all algorithms increases with the increase in the number of user antennas, but the computational complexity of the present invention is much smaller than that of the LS, LMMSE and KF algorithms, which can well save computing resources, reflecting the superiority of the present invention.
[0151] from Figure 4 It can be seen that under any SNR conditions, the performance of the scheme proposed in the present invention is better than that of the LS and LMMSE algorithms. Although the performance is weaker than that of the KF algorithm, the computational complexity of the scheme of the present invention is the lowest, and better tracking estimation performance can be achieved with extremely low computational complexity when the scatterer changes suddenly.
[0152] from Figure 5 It can be seen that with the increase of the number of time slots t, the performance of the scheme proposed in the present invention and the KF algorithm gradually decreases, but in the first 4 time slots, they are better than the LS and LMMSE algorithms, which verifies the feasibility of the scheme of the present invention in tracking angle parameters in multiple time slots.
[0153] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should be included in the scope of the claims of the present invention. The techniques, shapes, and structural parts not described in detail in the present invention are all known technologies.
Claims
1. A millimeter wave channel tracking method based on a reconfigurable smart surface, characterized in that: The method is applied to the millimeter wave channel tracking problem in an environment where scatterers suddenly change and the user end moves slowly, and includes the following steps: S1. Construct a RIS-assisted uplink wireless communication system model, as well as a dual-time scale model and an angle variation model of the UE-RIS-BS link; S2. Initialize the channel angle parameters using a two-dimensional fast Fourier transform algorithm and initialize the path gain parameters using a maximum likelihood algorithm; S3, tracking the channel angle in the horizontal direction of UE-RIS based on Newton algorithm; S4, detecting whether the channel matrix H(t) will undergo a mutation. If a mutation is detected in the channel matrix H(t), the parameters are initialized again. Otherwise, the Newton algorithm is used to continue tracking the angle parameters. The channel of the UE-RIS-BS link can only communicate through RIS end reflection, and the method of step S1 specifically includes the following steps: S101, equip BS with N R Uniform linear array antennas, serving N T For a single user with one antenna, the RIS is composed of Reflection units, among which, and are the number of reflection elements in the horizontal and vertical directions of RIS, respectively; S102, L radio frequency chains are connected in the horizontal and vertical directions of RIS, and the direct path from UE to BS is blocked. The signal sent by UE in the tth time slot can only be reflected to BS through RIS, and the received signal of BS for: Y(t)=GH(t)Γ(t)+N(t) in, is the channel matrix of RIS-BS, is the channel matrix of UE-RIS, is the pilot signal sent by the UE, T is the number of time slots, and the pilot signal meets the orthogonal condition N(t) has a mean of 0 and a variance of The probability distribution of complex Gaussian white noise is expressed as S103, assuming that the millimeter wave channel G of RIS-BS is a quasi-static channel, model the channel matrices G and H(t) of UE-RIS and RIS-BS respectively; The channel matrices G and H(t) are modeled based on the dual time scale model. In step S103, the UE-RIS and RIS-BS channel matrices are specifically modeled as follows: Among them, L a and L b are the number of paths between UE-RIS and RIS-BS, respectively. is the lth channel H(t) a The path gain, is the lth channel G b The path gain, is the UE Angle of Departure (AoD), and are the elevation and azimuth of the RIS angle of arrival (AoA), and are the elevation and azimuth of the AoD of RIS, is the AoA of BS, and is the array steering vector of channel H(t); Based on the Markov chain modeling of the angle change, the array steering vector of the channel H(t) and They are: in, d U and d I are the antenna distances between UE and RIS, respectively, and λ is the wavelength of electromagnetic waves; Expressions and Similar, and d U Use and d B Instead, d B is the BS antenna spacing, Expressions and Similar, and Use and replace; The angle parameter set O(t) of all UE-RIS paths in the tth time slot is: Since the channel angle change between adjacent time slots can be modeled as a first-order Markov chain, the angle model when there is an angle change in the t-th time slot is: O(t)=O(t-1)+u(t) in, Indicates that the mean is 0 and the variance of the angle change is The complex Gaussian white noise, Q U is the covariance matrix of u(t); In step S2, a two-dimensional fast Fourier transform algorithm is used to initialize the channel angle parameters, including RIS estimating the angle in the horizontal direction and RIS estimating the angle in the vertical direction. The method of estimating the angle in the horizontal and vertical directions of the RIS is the same. The RIS estimating the angle in the horizontal direction specifically includes the following steps: S201, RIS horizontal receiving signal for: in, is the horizontal channel matrix from UE to RIS, N x (t) is Gaussian white noise, and the received signal is decorrelated with the pilot signal and then filled with zero to obtain ; S202, obtaining the signal spatial spectrum through 2D-FFT Specifically: in, H x The spatial spectrum of (t), To obtain the Gaussian white noise N′ superimposed on the received signal after zero padding x The spatial spectrum of (t), F L′ and is the Fourier matrix after zero padding, the channel matrix The element in row m and column n of for: Among them, the definition After peak search, the l a The coordinates of the peak point of the path are get Therefore, the first horizontal direction from UE to RIS a The angle estimate for this path is: S203, perform 2D-FFT on the vertical direction received signal of the RF chain connected from the UE to the RIS to estimate for: in, for The corresponding peak point ordinate; In step S2, the maximum likelihood algorithm is used to initialize the path gain parameter, and the specific steps are as follows: Rewrite the signal model as: in, H x (t) Array steering matrix and the path gain matrix They are: Using ML algorithm to solve the estimated value of UE-RIS channel path gain parameter in the tth time slot for: in,(·) -1 To invert the matrix, and is the estimated value of the array steering matrix. Similarly, the RIS-BS channel path gain can be initialized; The method of step S3 specifically comprises the following steps: S301, after tracking and obtaining the angle parameter of a path, remove the received signal of the path to obtain the residual F x (t) is: in, and For the front a The estimated channel matrix parameters obtained by the iteration simplifies the residual energy R(t) to: Among them, l a =1,...,L a is the path number index, and The first a After iterations and estimated values The column vector of minimized residual energy R(t) is obtained and The value of can be used to trace the channel angle parameter value; S302, R(t) The Taylor expansion is: Among them, R u ′(t) and R v ′(t) is the first partial derivative of function R(t) with respect to u(t) and v(t), R u ″ u (t),R u ″ v (t),R v ″ u (t) and R v ″ v (t) are the second partial derivative of function R(t) with respect to u(t), the partial derivative of u(t) followed by the partial derivative of v(t), the partial derivative of v(t) followed by the partial derivative of u(t), and the second partial derivative of v(t); S303, transform the UE-RIS channel angle parameter tracking problem into solving R u ′(u(t),v(t))=0 and R v The problem of (u(t), v(t)) value when ′(u(t), v(t)) = 0, R u ′(u(t),v(t)) and R v ′(u(t),v(t)) is: S304, solve R by combining u ′(u(t),v(t))=0,R v ′(u(t),v(t))=0, we get: Because in the l a The channel angle parameter value obtained by looping only once in a tracking operation may not be the most accurate, so multiple loops are required. Assuming that loop R c Then the first a The channel angle parameter value obtained in the kth cycle of tracking is: Where k = 1,…,R c , is the first-order partial derivative, is the second-order partial derivative; S305, iteratively tracking the channel angle parameter value, the tracking value convergence condition is that the residual is less than the required accuracy value, and after the angle parameter value is tracked, the path gain parameter value is estimated using ML; S306: Use the channel angle parameter value and the path gain parameter value to restore the channel matrix H of the UE-RIS horizontal RF chain in the tth time slot x (t).
2. The millimeter wave channel tracking method based on a reconfigurable smart surface according to claim 1, characterized in that: The method of step S4 specifically comprises the following steps: S401. Define the likelihood function as: Among them, Q U is the noise covariance matrix, To calculate the matrix Q U Determinant N T When the H0 event occurs, Approaching 0; S402. According to the Newman-Pearson criterion, the log-likelihood ratio is: When the judgment is an H1 event, the following conditions must be met: Where Υ is the predefined decision threshold, and the false alarm probability P of an H0 event being judged as an H1 event is preset. fa for: Given the false alarm probability P fa The value of the threshold Υ can be determined. It follows the chi-square distribution, defined as is the right tail probability, where p(t) is the probability density function, so the calculation threshold Υ is: S403, according to the decision criteria in step S402, when it is determined to be an H1 event, it can be considered that the channel has a sudden change, at this time the channel matrix is reinitialized, and then the next tracking process begins.
Citation Information
Patent Citations
Angle domain channel tracking method under large-scale MIMO high-speed moving scene
CN108494449A
Channel tracking method for FDD system
CN110430147A
Estimation and prediction method of RIS auxiliary channel based on a ZC sequence
CN113472705A
Sparse cascade channel estimation method for millimeter wave IRS cooperative system
CN114285702A
Low-complexity channel estimation method for reconfigurable intelligent surface-assisted millimeter wave communication
CN114520699A
