Double-branch compression CSI feedback method in UAV system

Abstract

The invention discloses a double-branch CSI feedback compression method in a UAV (unmanned aerial vehicle) system, which comprises the following steps: the UAV extracts sensing information according to a GNSS (global navigation satellite system) chip to obtain maximum delay expansion and maximum Doppler frequency shift; the UAV performs two-step error suppression based on the sensing information: initial error suppression and threshold-based error suppression; the UAV designs a double-branch compression feedback scheme based on the sensing information: a CF branch and an EF branch to obtain a final feedback vector; and the BS designs a double-branch reconstruction algorithm for different branches to obtain reconstructed CSI. According to the invention, the precision of downlink CSI is improved at a feedback source, and the feedback overhead is reduced by using a double-branch compression scheme; according to the method and the device, the CSI is optimized by the UAV by utilizing the sensing information, so that the calculation complexity of CSI reconstruction is reduced under the condition that the BS does not increase extra resource consumption, and the recovery precision of the CSI is improved.

Description

Technical Field

[0001] The present invention relates to a dual-branch compressed channel state information (CSI) feedback method in a UAV system, which is a CSI feedback technology in a massive multiple-input and multiple-output (mMIMO) communication system in a frequency division duplex (FDD) mode assisted by an unmanned aerial vehicle (UAV). Background Art

[0002] In the field of wireless communications, the mMIMO system in FDD mode has attracted widespread attention because of its ability to significantly improve spectrum efficiency and data transmission rate. However, in order to fully utilize the advantages of the mMIMO system, it is crucial to accurately feed back the CSI from the receiver to the transmitter. The accuracy of CSI feedback directly affects the performance of the system, including signal reliability and transmission efficiency. Traditional CSI feedback methods usually require users (UE, user equipment) to collect channel information and send it back to the base station (BS, base station) through limited feedback bandwidth. In mMIMO systems, this method faces multiple challenges, such as high CSI data dimension and large feedback overhead. To address these problems, researchers have proposed a variety of CSI compression and quantization techniques to reduce the amount of data that needs to be fed back. However, these methods often require complex calculations on the UE, which undoubtedly increases the energy consumption and processing burden of the UE, especially when resources and energy are limited.

[0003] In recent years, with the rapid development of wireless communication technology, integrated sensing and communication (ISAC) technology has gradually become a research hotspot as an emerging communication mode. In a sensing-assisted communication environment, UAVs can be flexibly deployed in the communication network and collect environmental information such as location and speed through their onboard sensors. This versatility brings new possibilities to wireless communication systems, especially in improving the acquisition and utilization efficiency of CSI. To this end, we can use the sensing information collected by UAVs to assist in compressing CSI, which can not only significantly reduce the feedback overhead, but also reduce the computational burden of UAVs, thereby providing a better CSI feedback solution for mMIMO systems in FDD mode. Summary of the invention

[0004] The research purpose of the present invention is to overcome the shortcomings of the prior art and propose a dual-branch compressed CSI feedback method in a UAV system. The present invention cleverly uses the perception information collected by the UAV to assist in compressing the CSI. The specific steps include: extracting the perception information, suppressing CSI errors based on the perception information, and designing a dual-branch CSI compression feedback method using the perception information. Compared with the traditional CSI feedback method, the present invention uses the perception information at the feedback source to improve the accuracy of the downlink CSI, and designs a dual-branch compression scheme to reduce the feedback overhead. At the same time, due to the optimized processing of the CSI by the UAV end, the BS reduces the computational complexity of CSI reconstruction without increasing additional resource consumption, and improves the recovery accuracy of the CSI.

[0005] The present invention is achieved through the following technical solutions:

[0006] A dual-branch compressed CSI feedback method in a UAV system comprises the following steps:

[0007] S1, based on the maximum perceived delay expansion and the maximum Doppler shift υ m , suppress the CE error of UAV and obtain the CSI matrix after error suppression

[0008] S2. Design noise reduction detection threshold T h , perform threshold-based noise reduction processing to obtain the noise reduction

[0009] S3, cropping noise reduction CSI matrix Get the trimmed CSI matrix

[0010] S4, according to the clipping CSI matrix Design compression feedback (CF) and extraction feedback (EF) dual branches to obtain feedback CSI vector

[0011] The CF branch refers to compressing the cropped CSI matrix again and then feeding it back;

[0012] The EF branch refers to the non-zero elements in the feedback clipping CSI matrix;

[0013] S5, gBS receives the CSI vector Reconstruct the downlink

[0014] Furthermore, the step S1 includes the following sub-steps:

[0015] S1.1. Perception prior information: According to the instantaneous velocity v perceived by the UAV at time t, the maximum Doppler frequency shift υ is obtained m :

[0016]

[0017] At the same time, according to the perceived straight-line distance R between the UAV and the BS and the Friis transmission formula, the maximum delay extension is obtained.

[0018]

[0019] Among them, f c is the downlink carrier frequency, c is the speed of light, is the propagation distance of the LoS path under a given transmit power and the minimum resolvable receive power;

[0020] S1.2, Normalization processing: υ m , Normalized to the DD domain grid, they are expressed as:

[0021]

[0022] in, and are the Doppler and delay resolutions, and T are the number of OFDM symbols and the symbol duration along the delay axis, respectively. and Δf represent the number of symbols and subcarrier spacing along the Doppler axis, respectively;

[0023] S1.3. DD domain grid division: Based on the prior information K m and L m , divide the DD domain grid into two regions and

[0024]

[0025] S1.4、Suppress CE error: Set the elements in to zero to suppress CE errors and obtain the CSI matrix

[0026]

[0027] Specifically, the threshold-based noise reduction process described in step S2 includes:

[0028] S2.1. Preliminary cropping matrix To reduce the complexity of threshold detection, the pruned CSI matrix is ​​obtained

[0029]

[0030] in,

[0031] S2.2. Noise regions identified based on perceptual information Design the noise reduction detection threshold T h :

[0032]

[0033] in, is the noise variance, P f is the constant false alarm probability;

[0034] S2.3, based on threshold T h Further noise reduction, the matrix Its (m,n)th element is expressed as:

[0035]

[0036] Furthermore, the dual-branch CSI compression method described in step S4 includes the following sub-steps:

[0037] S4.1. Calculate the feedback overhead D of CF and EF branches respectively CF and D EF , if D EF ≥D CF , then go to the CF branch described in step S4.2; otherwise go to the EF branch described in step S4.3;

[0038] S4.2. Execute the CF branch, including the following sub-steps:

[0039] (1) Vectorize the CSI matrix to obtain s′ CF :

[0040]

[0041] (2) Using the encoder to s′ CF Further compression to reduce feedback overhead, the compressed vector

[0042]

[0043] Among them, f en (·) represents the mapping function of the encoder, and Ψ represents the relevant parameters that the encoder needs to learn;

[0044] (3) Combination and the trimmed CSI matrix dimension indicator vector Υ=(P,Q)T , branch instruction I (I = 0, 1), and obtain the final feedback vector s of the CF branch CF :

[0045]

[0046] S4.3. Execute the EF branch, including the following sub-steps:

[0047] (1) Obtaining the CSI Matrix The support set

[0048]

[0049] (2) According to the support set Extract the CSI matrix sequentially The non-zero elements in

[0050]

[0051] (3) Combine the gains of all antennas With support set Forming CSI vector

[0052]

[0053] (4) Combination With the branch instruction I, the final feedback vector s of the EF branch is obtained EF :

[0054]

[0055] Furthermore, the dual-branch CSI reconstruction method described in step S5 includes the following sub-steps:

[0056] S5.1. CSI vector received from BS Extract branch instruction I, if I = 1, then go to step S5.2 to execute CF reconstruction branch; if I = 0, then go to step S5.3 to execute EF reconstruction branch;

[0057] S5.2. Execute the CF reconstruction branch, including the following sub-steps:

[0058] (1) Using the decoder Decode and get the decoded vector s′ CF :

[0059]

[0060] Among them, f de(·) is the mapping function of the decoder;

[0061] (2) According to the decoded vector s′ CF Forming a single antenna CSI matrix Its (i,j)th element is expressed as:

[0062]

[0063] (3) The single antenna CSI matrix is ​​filled with zeros Restore to original matrix dimensions

[0064]

[0065] S5.3. Execute the EF reconstruction branch, including the following sub-steps:

[0066] (1) According to the feedback vector The significant difference between the gain and support data is extracted and sparsity

[0067] (2) Based on and Reconstructing the CSI matrix Its (i,j)th element is expressed as:

[0068]

[0069] in, and Respectively and No. elements;

[0070] S5.4, through the symplectic finite inverse Fourier transform, and Convert from DD domain to TF domain:

[0071]

[0072] S5.5. Collect the CSI matrices of all antennas to form the final recovered CSI matrix Its nth t Matrix It is expressed as:

[0073] BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0075] Figure 2 is a flowchart of the error suppression process based on perception information of the present invention;

[0076] Figure 3 This is a flow chart of the dual-branch compression feedback processing of the present invention. DETAILED DESCRIPTION

[0077] The present invention is described in detail below in conjunction with the embodiments and drawings, but it should be understood that the embodiments and drawings are only used to exemplify the present invention and do not constitute any limitation on the protection scope of the present invention. All reasonable changes and combinations within the scope of the inventive concept of the present invention fall within the protection scope of the present invention.

[0078] Reference Figure 1 , a dual-branch compressed CSI feedback method in a UAV system includes:

[0079] S1. UAV extracts perception information from the GNSS chip and processes it to obtain the maximum delay extension and the maximum Doppler shift υ m , these two parameters correspond to the grid value L according to the grid resolution m and K m ;

[0080] Among them, some more specific implementation methods are as follows:

[0081] S2, UAV performs two-step error suppression based on the perception information, including initial error suppression and threshold-based error suppression. Specifically, according to the perception information L m and K m , the area Set the elements in zero to complete the initial error suppression, and get Based on The element design detection threshold T h , and error suppression is performed again to obtain

[0082] Among them, some more specific implementation methods are as follows:

[0083] S3, UAV designs a dual-branch compression feedback scheme based on perception information, including CF branch and EF branch. By comparing the feedback overhead D of the two branches CF and D EF , select to enter the corresponding compressed branch.

[0084] The CF branch includes: firstly, vectorizing the CSI matrix to obtain s' CF ; then compress s' through the encoder CF ,get Finally, the CF feedback vector s is formedCF ;

[0085] The EF branch includes: obtaining the support set of the CSI matrix And according to the support set Extract the non-zero elements and get Finally, the EF feedback vector s is formed EF ;

[0086] Among them, some more specific implementation methods are as follows:

[0087] S4. BS designs a corresponding dual-branch reconstruction algorithm to restore the CSI matrix. According to the branch instruction I, the compression branch taken by the UAV is identified and the corresponding reconstruction branch is entered.

[0088] The CF reconstruction branch includes: extracting relevant parameters, decoding the compressed vector, and reshaping the decoded vector into a CSI matrix;

[0089] The EF reconstruction branch includes: extracting relevant parameters and restoring the path information of the CSI matrix through mapping.

[0090] Reference Figure 2 , step S2 further includes the following sub-steps:

[0091] S2.1. Division of regions and Based on the perception information K m and L m Divide the DD domain grid area into: path area and off-path areas Right now:

[0092]

[0093] S2.2, Initial error suppression: Non-path area Set the elements in zero to obtain the CSI matrix after initial error suppression

[0094]

[0095] At the same time Crop (retain only region), get Facilitates subsequent compression of CSI;

[0096] S2.3, Threshold-based Error Suppression: According to The distribution characteristics of elements in the area and the design of the detection threshold T h :

[0097]

[0098] Then error suppression is performed again based on the threshold to obtain the matrix

[0099]

[0100] S2.4, Crop the CSI matrix: If the CSI matrix after threshold error suppression If the value contains zeros for the entire row or column, it will be further pruned to obtain the CSI matrix

[0101]

[0102] Reference Figure 3 , step S3 further includes the following sub-steps:

[0103] S3.1. Calculate the feedback cost D of CF and EF branches CF and D EF :

[0104]

[0105] If D CF ≤D EF , then go to step S3.2 and execute CF branch; if D CF >D EF , then go to step S3.3 to execute the EF branch;

[0106] S3.2, execute the CF branch, including the following sub-steps:

[0107] (1) Vectorize the CSI matrix to obtain s′ CF :

[0108]

[0109] (2) Using the encoder to s′ CF Further compression to reduce feedback overhead, the compressed vector

[0110]

[0111] Among them, f en (·) represents the mapping function of the encoder, and Ψ represents the relevant parameters that the encoder needs to learn;

[0112] (3) Combination and dimension indicator vector Υ=(P,Q) T , branch instruction I (I = 0, 1), and obtain the final feedback vector s of the CF branch CF :

[0113]

[0114] S3.3, executing the EF branch, including the following sub-steps:

[0115] (1) Obtaining the CSI Matrix The support set

[0116]

[0117] (2) According to the support set Extract the CSI matrix sequentially The non-zero elements in

[0118]

[0119] (3) Combine the gains of all antennas With support set Forming CSI vector

[0120]

[0121] (4) Combination Together with the branch instruction I, it forms the final feedback vector s of the EF branch. EF :

[0122]

[0123] Example 1

[0124] In step S2, based on the perception information, the nth t The CSI on the root antenna is error suppressed and the A specific embodiment is as follows:

[0125] Assume L = 8, M = 128, N = 32, P = 2, Q = 3, P f =1e-6, Generated perfect channel express:

[0126]

[0127] Downlink CSI matrix after initial error suppression for:

[0128]

[0129] in, is the CSI matrix after initial cropping, expressed as:

[0130]

[0131] The CSI matrix after threshold detection and error suppression for:

[0132]

[0133] It can be seen that the matrix contains zero values ​​for the entire row or column. The matrix is ​​clipped again to obtain

[0134]

[0135] The above embodiments are only preferred implementations of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A dual-branch compressed CSI feedback method in a UAV system, characterized in that: The following steps are involved: S1, based on the maximum perceived delay expansion and the maximum Doppler shift υ m , suppress the CE error of UAV and obtain the CSI matrix after error suppression S2. Design noise reduction detection threshold T h , perform threshold-based noise reduction processing to obtain the noise reduction S3, cropping noise reduction CSI matrix Get the trimmed CSI matrix S4, according to the clipping CSI matrix Design the compression feedback CF and extraction feedback EF dual branches to obtain the feedback CSI vector The CF branch refers to compressing the cropped CSI matrix again and then feeding it back; The EF branch refers to the non-zero elements in the feedback clipping CSI matrix; S5, gBS receives the CSI vector Reconstruct the downlink 2. A dual-branch compressed CSI feedback method in a UAV system according to claim 1, characterized in that: Suppressing CE errors in step S1 includes the following sub-steps: S1.

1. Perception prior information: According to the instantaneous velocity v perceived by the UAV at time t, the maximum Doppler frequency shift υ is obtained m : At the same time, according to the perceived straight-line distance R between the UAV and the BS and the Friis transmission formula, the maximum delay extension is obtained. Among them, f c is the downlink carrier frequency, c is the speed of light, l LoS is the propagation distance of the LoS path under a given transmit power and the minimum resolvable receive power; S1.2, Normalization processing: υ m , Normalized to the DD domain grid, they are expressed as: in, and are the grid resolutions for Doppler and delay, respectively, and T are the number of OFDM symbols and the symbol duration along the delay axis, respectively. and Δf represent the number of symbols and subcarrier spacing along the Doppler axis, respectively; S1.

3. DD domain grid division: Based on the prior information K m and L m , divide the DD domain grid into two regions and S1.4、Suppress CE error: Set the elements in to zero to suppress CE errors and obtain the CSI matrix 3. The dual-branch compressed CSI feedback method in a UAV system according to claim 1, characterized in that: The threshold-based noise reduction process in step S2 includes the following sub-steps: S2.

1. Preliminary cropping matrix To reduce the complexity of threshold detection, the pruned CSI matrix is ​​obtained in, S2.

2. Noise regions identified based on perceptual information Design the noise reduction detection threshold T h : in, is the noise variance, P f is the constant false alarm probability; S2.3, based on threshold T h Further noise reduction, the matrix Its (m,n)th element is expressed as:

4. The dual-branch compressed CSI feedback method in a UAV system according to claim 1, characterized in that: The dual-branch compressed CSI feedback method in step S4 includes: S4.

1. Calculate the feedback overhead D of CF and EF branches respectively CF and D EF , if D EF ≥D CF , then go to the CF branch described in step S4.2; otherwise go to the EF branch described in step S4.3; S4.

2. Execute the CF branch, including the following sub-steps: (1) Vectorize the CSI matrix to obtain s′ CF : (2) Using the encoder to s′ CF Further compression to reduce feedback overhead, the compressed vector Among them, f en (·) represents the mapping function of the encoder, and Ψ represents the relevant parameters that the encoder needs to learn; (3) Combination and the trimmed CSI matrix dimension indicator vector Υ=(P,Q) T , branch instruction I, I = 0, 1, get the final feedback vector s of the CF branch CF : S4.

3. Execute the EF branch, including the following sub-steps: (1) Obtaining the CSI Matrix The support set (2) According to the support set Extract the CSI matrix sequentially The non-zero elements in (3) Combine the gains of all antennas With support set Forming CSI vector (4) Combination With the branch instruction I, the final feedback vector s of the EF branch is obtained EF :

5. The dual-branch compressed CSI feedback method in a UAV system according to claim 1, characterized in that: The dual-branch CSI reconstruction method described in step S5 includes the following sub-steps: S5.

1. CSI vector received from BS Extract branch instruction I, if I = 1, then go to step S5.2 to execute CF reconstruction branch; if I = 0, then go to step S5.3 to execute EF reconstruction branch; S5.

2. Execute the CF reconstruction branch, including the following sub-steps: (1) Using the decoder Decode and get the decoded vector s′ CF : Among them, f de (·) is the mapping function of the decoder; (2) According to the decoded vector s′ CF Forming a single antenna CSI matrix Its (i,j)th element is expressed as: (3) The single antenna CSI matrix is ​​filled with zeros Restore to original matrix dimensions S5.

3. Execute the EF reconstruction branch, including the following sub-steps: (1) According to the feedback vector The significant difference between the gain and support data is extracted and sparsity (2) Based on and Reconstructing the CSI matrix Its (i,j)th element is expressed as: in, and Respectively and No. elements; S5.4, through the symplectic finite inverse Fourier transform, and Convert from DD domain to TF domain: S5.

5. Collect the CSI matrices of all antennas to form the final recovered CSI matrix Its nth t Matrix It is expressed as:

Images