Channel knowledge base construction method based on environment topology and channel measurement sequence

Abstract

The invention discloses a channel knowledge base construction method based on environment topology and a channel measurement sequence. The method comprises the following steps: defining a wireless channel, a channel knowledge base and a channel observation model in an MIMO system; giving an environment topology and a channel measurement sequence, and carrying out initial user positioning based on the environment topology and the channel measurement sequence; based on environment topology and a channel measurement sequence, modeling a channel knowledge base construction problem, solving a user position sequence, and then solving a covariance matrix in a channel knowledge base # imgabs0 # to realize construction of the channel knowledge base; and iteratively solving the user position sequence and constructing the channel knowledge base to obtain an optimal user position sequence and channel knowledge base. According to the method, the channel knowledge base can be constructed only according to the environment topology and the low-dimensional channel observation sequence without an accurate user measurement position or accurate channel space feature data, so that the construction difficulty and cost of the channel knowledge base are effectively reduced.

Description

Technical Field

[0001] The invention relates to the construction of a channel knowledge base in a wireless network, and in particular to a channel knowledge base construction method based on environment topology and channel measurement sequence. Background Art

[0002] The channel knowledge base is a map that describes the spatial characteristics of the channel. It can map any spatial position coordinates to the spatial characteristics of the channel. These spatial characteristics include but are not limited to the channel covariance matrix, signal departure angle, signal arrival angle, received signal strength, signal-to-noise ratio, etc. The channel knowledge base can be used to analyze the spatial signal characteristics of the system, assist in network equipment deployment, trajectory planning, etc. The channel knowledge base can also be used to assist in channel estimation, beam tracking and user positioning.

[0003] Most existing methods for building channel knowledge bases require accurate location information and spatial characteristics of the channel at a specific location. For example, in a single-input-single-output (SISO) system, a received signal strength map can be established by measuring the received signal strength at each specific location. In a multi-input-multi-output (MIMO) system, a beam index map can be established by performing beam scanning at each specific location.

[0004] However, obtaining precise positions requires additional hardware and higher time costs, and it is also difficult to obtain complete channel spatial features. Especially in MIMO systems, the channels are usually of extremely high dimensions, and a large number of pilot signals are required to obtain accurate and complete channels. Even if there are a large number of pilot signals for channel perception, the extraction of accurate channel features is extremely challenging and is currently a hot topic in research.

[0005] In existing solutions, constructing radio spectrum maps or channel knowledge bases mostly requires accurate user sampling locations. However, obtaining accurate locations requires additional hardware equipment and higher time costs; constructing radio spectrum maps or channel knowledge bases mostly requires accurate channel spatial features, such as channel arrival angles and high-dimensional channel vectors. Especially in MIMO systems, channels usually have extremely high dimensions. To obtain accurate and complete channels, a large number of pilot signals are required. Even if there are a large number of pilot signals for channel perception, it is difficult to extract accurate channel features. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for constructing a channel knowledge base based on environmental topology and channel measurement sequence. It does not require precise user sampling positions or accurate channel spatial characteristics. The channel knowledge base can be constructed only according to the environmental topology and low-dimensional channel observation sequences, which effectively reduces the difficulty and cost of constructing the channel knowledge base.

[0007] The object of the present invention is achieved through the following technical solution: a method for constructing a channel knowledge base based on environment topology and channel measurement sequence, comprising the following steps:

[0008] Define wireless channels, channel knowledge base and channel observation model in MIMO systems;

[0009] Given an environment topology and a channel measurement sequence, initial user positioning is performed based on the environment topology and the channel measurement sequence;

[0010] Based on the environment topology and channel measurement sequence, the channel knowledge base construction problem is modeled and the user position sequence is solved. The covariance matrix in is solved to build the channel knowledge base;

[0011] The user position sequence is solved and the channel knowledge base is constructed iteratively to obtain the optimal user position sequence and channel knowledge base.

[0012] The beneficial effects of the present invention are as follows: the present invention preliminarily and roughly estimates the user position based on the base station position and the signal strength of the channel observation, constructs a hidden Markov problem based on the roughly estimated user position, low-dimensional (compared to high-dimensional channel) channel measurement, channel knowledge base (randomly initialized in the first iteration) and user movement statistical characteristics, and uses the backtracking algorithm to solve the optimal trajectory sequence; constructs (or updates) the channel knowledge base according to the solved trajectory sequence and low-dimensional channel measurement. (The first iteration is to construct the channel knowledge base, and the second iteration starts to update the channel knowledge base); repeats the iteration to achieve the construction of the channel knowledge base; no precise user sampling position or accurate channel spatial characteristics are required, and the channel knowledge base can be constructed only according to the environmental topology and the low-dimensional channel observation sequence, which effectively reduces the difficulty and cost of constructing the channel knowledge base. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the principle of the present invention;

[0014] Figure 2 Schematic diagram of channel estimation performance driven by the reconstructed channel knowledge base. DETAILED DESCRIPTION

[0015] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.

[0016] like Figure 1 As shown, a method for constructing a channel knowledge base based on environment topology and channel measurement sequence includes the following steps:

[0017] Define wireless channels, channel knowledge base and channel observation model in MIMO systems;

[0018] The technical solution of the present invention is illustrated by taking a wireless channel in a multiple-input multiple-output (MIMO) network as an example. The channel knowledge base in this example is visualized as a mapping from spatial position to channel covariance matrix, but the channel spatial features carried by the channel knowledge base are not limited to the channel covariance matrix, but can also be variables such as signal departure angle, signal arrival angle, received signal strength, signal-to-noise ratio, etc. that are related to the spatial position of the signal transceiver device and can characterize a certain aspect of the channel.

[0019] Consider a MIMO system where each base station is equipped with massive MIMO antennas, with N t antenna units, and mobile users use a single antenna for communication.

[0020] First, define the wireless channel as h. Since the base station has N t antenna elements, so the channel h is an N t Since the channel is related to both time and space, let h be the channel at time t. t Modeled as a first-order autoregressive model:

[0021] h t =γh t-1 +(1-γ 2 ) t (1)

[0022] Among them, γ is the first-order autoregressive coefficient, h t-1 is the channel at time t-1, u t It is a vector related to the channel spatial statistical characteristics. Define the user's position at time t as p t , here we assume that u t ~CN(0,C(p t )), that is u t Subject to mean 0 and variance p t The channel covariance matrix C(p t ) complex Gaussian distribution, where c(p t ) is provided by the channel knowledge base. According to the above channel model (1), it can be deduced that: ht ~CN(0,C(p t )).

[0023] Define a channel knowledge base as

[0024]

[0025] Where x = (x1, x2, x3) T is a three-dimensional position vector, x is the set of all spatial grid center positions after discretizing the target area into equidistant grids, and C(x) is the average channel covariance matrix at the position in the grid with the center position x, which is defined as here Represents the covariance hh of all channels h in the grid H Taking the expectation, by definition, c(x) is an N t ×N t Dimensional matrix. Since the user position p t falls in the grid of the target area. If the center of the grid is x, then u in the channel model t The covariance C(p t ) can be replaced by C(x).

[0026] Given M pilot signals, we can observe y at time t t Create the following model:

[0027] y t =A t h t +n t (3)

[0028] Here, A t is a channel sensing matrix at the base station at time t, each row of which corresponds to a pilot signal, so A t There are M rows and N t Column, h t is an N t The column vector of rows is the real channel at time t, n t is the measurement noise at time t, which is a column vector with M rows. Assume That is n t Each element of is subject to a mean of 0 and a variance of Gaussian distribution, where I is a column vector with M rows. According to the above channel measurement model (3), it can be deduced that in Yes A t The conjugate transpose of .

[0029] definition For the observation sequence from t = 1 to t = T, define is the user trajectory from t = 1 to t = T, and we define is the channel sequence from t=1 to t=T.

[0030] Given an environment topology, user positioning is performed based on the environment topology;

[0031] Given an environment topology represented as θ, the environment topology here can be concretized as the base station locations b1, b2, ..., b Q , that is, θ=[b1,b2,...,b Q ]. The environment topology here can also include the location of mobile devices as reference points, etc.

[0032] The weighted centroid localization (WCL) method can be used to obtain an inaccurate user location based on the received signal strength. Specifically, given the channel observation of each base station at time t, in, For q∈{1, 2, ..., Q}, it represents the channel observation received by the qth base station at time t.

[0033] The signal strength received by the qth base station at time t can be expressed as Define the location weight of the qth base station as w q , then the weight can be defined as

[0034]

[0035] Based on the location of each base station, the user’s location p at time t t can be estimated as a function of the environment topology θ and the observation Function

[0036]

[0037] in, represents the user position at time t estimated in the initial stage.

[0038] Based on the environment topology and channel measurement sequence, the channel knowledge base construction problem is modeled and the user position sequence is solved. The covariance matrix in is solved to build the channel knowledge base;

[0039] Based on the above model, we define Channel measurement sequence Channel Sequence Trajectory sequence (user location sequence) The joint probability distribution of In the process of constructing the channel knowledge base, this technique only assumes that the channel measurement sequence and the environment topology θ are known, and the real channel sequence and user location sequence Unknown, these unknown quantities and known quantities need to use the channel knowledge base The channel space features in the channel are connected in series, that is, the user position indicates the channel space features (through the channel knowledge base The channel covariance matrix in ), the channel spatial characteristics reflect the spatial distribution characteristics of the real channel, and some characteristics of the real channel are captured by a small number of pilots and saved in the channel observation.

[0040] In summary, the channel knowledge base construction problem can be modeled as a sequence of user positions. The solution of The construction of the channel covariance matrix corresponding to each position in . The mathematical expression of this problem is:

[0041]

[0042] in, is a system that contains the environment topology θ, i.e., the set of base station locations, and signal observations The regularization term is used to provide additional information for solving more accurate user measurement trajectory sequences. Without this additional information, the only channel observation information It cannot directly indicate the user's position coordinates in the real physical space. μ is a regularization coefficient.

[0043] Next, the user position sequence is and channel knowledge base The covariance matrix in is solved.

[0044] 1) User location sequence Solution

[0045] According to the channel observation model (3), we can get It can be seen that the channel observation y t With user position p t is directly related, so in solving When the joint probability distribution The channel sequence in can be ignored. Using Bayesian theory, the logarithmic joint probability distribution in problem (4) is is derived as:

[0046]

[0047] in, is a given position p t When y is observed t The conditional probability is based on the channel knowledge base C(p t ), according to the channel observation model (3), From position p t-1 Transfer to p t Here, according to the channel model (1) and the channel observation model (3), we can deduce The transition probability of the user position p(p t |p t-1 ) and the initial position probability p1 can be assumed to be known because they can be estimated from the user's historical motion data. Note that the channel covariance matrix in the channel knowledge base here can be initialized as a unit diagonal matrix, but the initialization method is not limited to this, and random initialization can also be performed.

[0048] From (5), we can see that the user position sequence in problem (4) is The problem to be solved is a hidden Markov problem, which can be solved by the Viterbi algorithm. The Viterbi algorithm can efficiently traverse all possible position sequences to find the sequence that maximizes the objective function value (joint probability) and best matches the observation.

[0049] 2) Channel Knowledge Base Construction

[0050] Channel Knowledge Base The construction of is mainly to construct the channel covariance matrix C(x) in each grid of the target area. According to the solved user position sequence All channel observations can be assigned to each grid. The center position of the observation is x i The grid is defined For all positions p t The set of moments in this grid. Channel observation y t can be used to reconstruct the channel covariance matrix. Note that the observation y t The dimension is M, which may be much smaller than the dimension N of the channel t .

[0051] One feasible solution is to construct an observation sample covariance matrix Ω y (x i ), then based on Ω y (x i ) derives a channel covariance matrix c(x i). Specifically, the observed sample covariance matrix Ω y (x i ) can be defined as:

[0052]

[0053] here, is located at the center position x i The number of channel observations in the grid, represents a projection of the channel observation, which is an N t ×N t The matrix of .

[0054] Based on Ω y (x i ), the channel covariance matrix C(x i An unbiased estimate of ) is:

[0055]

[0056] in, I N is an N t The unit column vector of the rows.

[0057] Unbiased estimation of the channel covariance matrix based on each grid The channel knowledge base is constructed for each position x i The channel covariance matrix of the grid at this location The mapping of , that is, the constructed channel knowledge base is expressed as

[0058] The user position sequence is solved and the channel knowledge base is constructed iteratively to obtain the optimal user position sequence and channel knowledge base.

[0059] Based on the channel knowledge base constructed in The user position sequence can be solved again to obtain a more accurate user position sequence; further, using the more accurate user position sequence, the channel covariance estimation method can be used again to update the channel covariance and channel knowledge base. Repeat the above steps until the change in the user position sequence of two iterations is less than a certain threshold (preset threshold), that is, the optimal user trajectory sequence and channel knowledge base are obtained.

[0060] The following is a simulation performance of a specific embodiment of the present invention. This embodiment is deployed in a 740m×710m urban environment. The target area contains 7 base stations. Each base station is configured with an N t =MIMO antenna composed of 64 antenna units, and base stations are randomly deployed on the top of some buildings.

[0061] Consider a user moving on the road at a speed of 10 m / s and a pilot signal transmission interval of 100 ms. At each user location, using the received signal strength and the known base station location, a rough user location can be obtained using weighted centroid positioning technology. The distance between the rough user location and the actual user location follows a Gaussian distribution with a mean of 0 and a variance of 30 m. Figure 1 The performance of channel estimation using the reconstructed channel knowledge base is demonstrated, and its evaluation index is the optimal ratio of channel capacity, that is, the ratio of the channel capacity achieved based on the estimated channel at each position to the maximum channel capacity that can be achieved by a perfect channel. This embodiment builds a channel knowledge base based on 280,000 channel observations (each channel observation contains 16 pilots) and rough position information. After the channel knowledge base is built, it is used to perform channel estimation and tracking. In this process, each user position only needs to transmit one pilot signal.

[0062] from Figure 2 It can be seen that the channel estimation scheme based on the prior knowledge of the channel covariance matrix provided by the perfect channel knowledge base (user location is unknown) can achieve 97% of the performance of the optimal channel capacity under the condition of 20dB signal-to-noise ratio. The channel knowledge base reconstructed based on the environment topology and channel measurement sequence proposed in the present invention can obtain a performance exceeding 88% of the final channel capacity when the signal-to-noise ratio is 5dB by using the channel covariance matrix knowledge provided by the reconstructed channel knowledge base. Other comparative schemes include 1) a channel estimation scheme driven by a perfect channel knowledge base with known positions: assuming that the channel covariance matrix of each observation position is known, the channel is estimated using the channel covariance matrix and Kalman filtering technology; 2) a perfect channel knowledge base scheme: assuming that the observation position is unknown but the channel knowledge base is known, the observation position can be estimated using the Bayesian posterior probability, and then the channel is estimated using the channel covariance matrix and Kalman filtering technology at that position; 3) a channel estimation scheme driven by a channel angle map: this scheme requires the construction of a mapping map from the position vector to the signal arrival angle, and uses the channel angle map and the channel observation at the current moment to perform channel estimation; 4) a Kalman filtering scheme: this scheme uses traditional Kalman filtering technology for channel estimation and tracking.

[0063] The above is a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not depart from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for constructing a channel knowledge base based on environment topology and channel measurement sequence, characterized in that: The following steps are involved: Define wireless channels, channel knowledge base and channel observation model in MIMO systems; Given an environment topology and a channel measurement sequence, initial user positioning is performed based on the environment topology and the channel measurement sequence; Based on the environment topology and channel measurement sequence, the channel knowledge base construction problem is modeled and the user position sequence is solved. The covariance matrix in is solved to build the channel knowledge base; The user position sequence is solved and the channel knowledge base is constructed iteratively to obtain the optimal user position sequence and channel knowledge base.

2. The method for constructing a channel knowledge base based on environment topology and channel measurement sequence according to claim 1, characterized in that: Defining the wireless channel, channel knowledge base and channel observation model in the MIMO system includes: Assume that in the MIMO system, each base station is equipped with a large-scale MIMO antenna, which has N t antenna units, mobile users use a single antenna for communication; Define the wireless channel as h: Since the base station has N t antenna units, so the channel h is an N t Dimensional column vector, since the channel is related to both time and space, the channel h at time t t Modeled as a first-order autoregressive model: h t =γh t-1 +(1-γ 2 )u t (1) Among them, γ is the first-order autoregressive coefficient, h t-1 is the channel at time t-1, u t It is a vector related to the channel spatial statistical characteristics; the user's position at time t is defined as p t , assuming u t ~CN(0,C(p t )), that is u t Subject to mean 0 and variance p t The channel covariance matrix C(p t ) complex Gaussian distribution, where C(p t ) is provided by the channel knowledge base; according to the channel model (1), it is derived that: h t ~CN(0,C(p t )); Define a channel knowledge base as Among them, x = (x1, x2, x3)w is a three-dimensional position vector, is the set of all spatial grid center positions after discretizing the target area into equidistant grids. C(x) is the average channel covariance matrix at the position in the grid with the center position x, which is defined as Represents the covariance hh of all channels h in the grid H Taking the expectation, C(x) is an N t ×N t Dimensional matrix, due to the user position p t falls in the grid of the target area. If the center of the grid is x, then u in the channel model t The covariance C(p t ) is replaced by C(x); Given M pilot signals, the channel observation y at time t is t Create the following model: y t =A t h t +n t (3) Among them, A t is a channel sensing matrix at the base station at time t, each row of which corresponds to a pilot signal. t There are M rows and N t Column, h t is an N t The column vector of rows is the real channel at time t, n t is the measurement noise at time t, which is a column vector with M rows; assuming That is n t Each element of has a mean of 0 and a variance of Gaussian distribution, where I is a column vector of M rows; according to the above channel measurement model (3), we can deduce in Yes A t The conjugate transpose of ; definition The observation sequence from t=1 to t=T defines the user position sequence Contains the user trajectory from t = 1 to t = T, and defines is the channel sequence from t=1 to t=T.

3. The method for constructing a channel knowledge base based on environment topology and channel measurement sequence according to claim 1, characterized in that: The given environment topology and channel measurement sequence, and performing initial user positioning based on the environment topology and the channel measurement sequence includes: Given an environment topology denoted as θ, the environment topology is concretized as the base station locations b1, b2, ..., b Q , that is, θ=[b1,b2,...,b Q ], The weighted centroid positioning method is used to obtain an inaccurate user location based on the received signal strength: Given the channel observation of each base station at time t in, For q∈{1, 2, ..., Q}, it represents the channel observation received by the qth base station at time t; where Q is the number of base stations; The signal strength received by the qth base station at time t is expressed as Define the location weight of the qth base station as w q , then the weight is defined as: Based on the location of each base station, the user’s location p at time t t is estimated as a function of the environment topology θ and the observation Function Right now: in, represents the user position at time t estimated in the initial stage.

4. The method for constructing a channel knowledge base based on environment topology and channel measurement sequence according to claim 1, characterized in that: The modeling of the channel knowledge base construction problem based on the environment topology and the channel measurement sequence includes: definition Channel measurement sequence Channel Sequence Trajectory sequence The joint probability distribution of In the process of building the channel knowledge base, it is assumed that the channel measurement sequence and the environment topology θ are known, and the real channel sequence and user location sequence Unknown, these unknown quantities and known quantities need to use the channel knowledge base The channel space features in the channel knowledge base are connected in series. The channel covariance matrix in ; The channel knowledge base construction problem is modeled as a sequence of user positions. The solution of The construction of the channel covariance matrix corresponding to each position in the channel knowledge base is mathematically expressed as: in, is a system that contains the environment topology θ, i.e., the set of base station locations, and signal observations The regularization term is used to provide additional information for a more accurate solution of the user measurement trajectory sequence. μ is a regularization coefficient.

5. The method for constructing a channel knowledge base based on environment topology and channel measurement sequence according to claim 4, characterized in that: The solving of the user position sequence comprises: According to the channel observation model (3), Channel observation y t With user position p t is directly related, so in solving When the joint probability distribution The channel sequence in can be ignored. Using Bayesian theory, the logarithmic joint probability distribution in problem (4) It is derived as follows: in, is a given position p t When y is observed t The conditional probability is based on the channel knowledge base C(p t ), according to the channel observation model (3), From position p t-1 Transfer to p t , p(p1) is the probability that the user is at position p1 at time t=1; According to the channel model (1) and the channel observation model (3), we can deduce The transition probability of the user position p(p t |p t-1 ) and the initial position probability p(p1) can be assumed to be known, which are estimated from the user's historical motion data; the channel covariance matrix in the channel knowledge base here is initialized to a unit diagonal matrix or randomly initialized. From (5), we can get the user location sequence in problem (4): The problem to be solved is a hidden Markov problem; the user position sequence is obtained by solving it through the classic Viterbi algorithm 6. The method for constructing a channel knowledge base based on environment topology and channel measurement sequence according to claim 5, characterized in that: The channel knowledge base The covariance matrix in is solved to build the channel knowledge base, including: Channel Knowledge Base The construction of the channel covariance matrix C(x) in each grid of the target area is mainly based on the solved user position sequence Allocate all channel observations to each grid; The center position of the survey is x i The grid is defined For all positions p t The set of moments in this grid; for Channel observation y t Used to reconstruct the channel covariance matrix and observe y t The dimension is M, which is much smaller than the dimension N of the channel t ; Construct an observation sample covariance matrix Ω y (x i ), then based on Ω y (x i ) derives a channel covariance matrix C(x i ) is an unbiased estimate of: Observation sample covariance matrix Ω y (x i ) is defined as: in, is located at the center position x i The number of channel observations in the grid, represents a projection of the channel observation, which is an N t ×N t Matrix of Based on Ω y (x i ), the channel covariance matrix C(x i An unbiased estimate of ) is: in, I N is an N t The unit column vector of the row; Unbiased estimation of the channel covariance matrix based on each grid The channel knowledge base is constructed for each position x i The channel covariance matrix of the grid at this location The mapping of , that is, the constructed channel knowledge base is expressed as 7. The method for constructing a channel knowledge base based on environment topology and channel measurement sequence according to claim 6, characterized in that: The iterative solution of the user trajectory sequence and the construction of the channel knowledge base include: Based on the channel knowledge base constructed in Channel Knowledge Base In As the grid center is x i The channel covariance matrix of the grid is t , the channel covariance matrix C(p t ) with Solve the user location sequence for problem (4) again to obtain a more accurate user location sequence; Using more accurate user position sequences, the channel covariance and channel knowledge base are updated again based on the channel covariance estimation method. Repeat the above steps until the change in the user position sequence between two adjacent iterations is less than a preset threshold, that is, the optimal user position sequence and channel knowledge base are obtained.

Images