A 6G Cell-free Massive MIMO Channel Calibration Method Based on Spanning Tree Algorithm
By applying the spanning tree algorithm in a large-scale MIMO system without cellular, a directed tree structure channel calibration relationship framework is built, which solves the problem of low channel calibration accuracy caused by the increase in the number of AP antennas, and achieves higher channel calibration accuracy and better Low computing complexity improves system performance.
Patent Information
- Application Number
- CN202211127541.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-16
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-09-16
AI Technical Summary
In cellular-free large-scale MIMO systems, the sharp increase in the number of AP antennas poses a huge challenge to the estimation problem of transmit channel state information (CSIT) at the base station, especially because the transmission overhead of wireless channels is too large, and the UE cannot transfer downlink CSIR. The estimation results are fed back to the AP, resulting in low channel calibration accuracy and affecting system performance.
A 6G cellular-free large-scale MIMO channel calibration method based on spanning tree algorithm is proposed. By constructing a wireless communication system in time division duplex TDD mode, a two-way service channel model between AP and UE and a two-way internal between any two APs are established. The channel model uses the directed tree structure of the graph theory model to build a channel calibration relationship framework, and minimizes the expected value of the system's relative error through the spanning tree algorithm, optimizes the topological relationship of channel calibration, and calculates the channel calibration coefficients of each AP under the cellular system.
It improves channel calibration accuracy, is better than traditional algorithms, has the lowest computing complexity, and significantly improves the performance of 6G cellular-free large-scale MIMO systems.
Smart Images

Figure CN115622639B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of 6G non-cellular massive MIMO channel calibration, and in particular to a non-cellular massive MIMO channel calibration method based on a spanning tree algorithm. Background Art
[0002] The cell-free massive MIMO wireless communication system has become one of the most promising technologies in 6G wireless communication technology. Benefiting from the characteristics of massive MIMO, the communication system can obtain spatial diversity gain without increasing the total transmission power of the system to improve energy efficiency and spectrum efficiency. Cell-free massive MIMO is a realization of the concept of scalable MIMO network and distributed multi-point cooperative network. The communication system does not have the concept of cell boundaries in traditional communications, so it effectively solves the boundary effect and further improves the performance of the system. The cell-free wireless communication system consists of a large number of distributed APs. All distributed APs simultaneously serve all users whose number is less than the number of AP antennas under the same time-frequency resource block. Each AP in the wireless communication system is connected to the central processing unit (CPU) through a backhaul link and transmits data to the UE through beamforming technology. Since the cell-free network can maximize the spatial diversity, its communication performance is better than that of traditional cellular cells.
[0003] In order to obtain better beamforming performance, accurate uplink and downlink channel estimation is necessary. However, in non-cellular massive MIMO systems, the sharp increase in the number of AP antennas poses a huge challenge to the transmit channel state information (CSIT) estimation problem at the base station. Specifically, during the wireless channel transmission process, due to the large transmission overhead of the wireless channel, the UE cannot feed back the estimated result of the downlink CSIR to the AP. Therefore, to solve this problem, the existing communication system adopts the TDD communication mode, using the reciprocity of the uplink and downlink channels, and uses the estimated result of the uplink CSIR at the base station as the downlink CSIT. The UE no longer needs to feed back the channel estimation result to each AP node, thereby saving communication overhead.
[0004] The uplink and downlink physical wireless channels between the AP and the UE are reciprocal, but a complete uplink and downlink channel includes not only the physical channel, but also the radio frequency (RF) link, digital-to-analog / analog-to-digital (DA / AD) converter, filter, mixer and antenna of the transceiver. The introduction of these hardware devices makes the entire wireless channel non-reciprocal. The channel calibration accuracy in the wireless communication system directly affects the upper bound of the system performance. Therefore, designing an accurate channel calibration method can greatly improve the system performance of 6G non-cellular massive MIMO. Summary of the invention
[0005] In view of this, the present invention considers the characteristics of distributed APs in non-cellular wireless communication from the perspective of improving the accuracy of channel calibration, and proposes a 6G non-cellular massive MIMO channel calibration method based on a spanning tree algorithm. The channel calibration process between M APs in the non-cellular communication system is modeled as a directed tree topology graph, and the expectation of minimizing the relative calibration error is taken as the optimization goal. A channel calibration method with low implementation complexity is proposed, and its performance is better than that of traditional algorithms.
[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:
[0007] A 6G non-cellular massive MIMO channel calibration method based on a spanning tree algorithm of the present invention comprises the following steps:
[0008] Step S1, constructing a 6G non-cellular massive MIMO wireless communication system in time division duplex (TDD) mode, establishing a two-way service channel model between AP and UE and a two-way internal channel model between any two APs, both of which are collectively referred to as a two-way wireless channel model, and the channel distortion factors considered include large-scale fading, small-scale fading and additive white Gaussian noise;
[0009] Step S2: A complete bidirectional channel includes not only the physical channel but also the RF link of the transceiver. The introduction of the RF link of the transceiver causes the entire wireless channel to be non-reciprocal. Aiming at the channel non-reciprocity problem caused by RF link mismatch, an air interface channel calibration problem in a non-cellular massive MIMO scenario is proposed with the goal of minimizing the expected value of the relative error of the channel calibration system.
[0010] Step S3: According to the air interface channel calibration problem described in step S2, a channel calibration relationship framework is constructed. Based on the large-scale fading prior information between channels within the AP, a directed tree structure of a graph theory model is used to construct a description of the topological calibration relationship between APs in the non-cellular communication system: the reference AP specified by the system is used as the root node of the channel calibration tree, and other APs are leaf nodes of the channel calibration tree. The directed edges of the channel calibration tree represent the calibration calculation relationship between APs.
[0011] Step S4, a spanning tree algorithm is proposed to minimize the expected value of the system relative error; in the proposed algorithm, the traditional Argos calibration topology structure is used as the initialization of the channel calibration tree, and the topological relationship of the channel calibration is optimized by modifying the directed connection relationship between the nodes in the channel calibration tree. The optimal channel calibration tree is obtained by continuous iteration and calculation, and the channel calibration coefficient of each AP in the non-cellular system is calculated based on the channel calibration tree.
[0012] Furthermore, the 6G non-cellular massive MIMO wireless communication system in the time division duplex TDD mode is constructed in step 1, and a bidirectional wireless channel model between the AP and the UE and between any two APs is established, specifically including:
[0013] Step S101: setting a distribution area of a non-cellular communication scenario, setting the number of distributed APs and UEs and their distribution locations, and establishing a communication model, specifically including:
[0014] In the Cartesian coordinate system, M APs are set with uniformly distributed two-dimensional coordinate positions in the non-cellular communication scene area, and the set of all APs is defined as
[0015] In the Cartesian coordinate system, K UEs are set with uniformly distributed two-dimensional coordinate positions in the non-cellular communication scenario area, and the set composed of all UEs is defined as
[0016] Step S102: Considering the large-scale fading and small-scale fading of the communication channel, the bidirectional service physical channel between the AP and the UE is modeled as:
[0017]
[0018] In the formula, β mk Indicates the label is The AP is labeled Large-scale fading between UEs, h mk Indicates the label is The AP is labeled Similarly, the bidirectional internal physical channel between any two APs labeled i and j is modeled as
[0019]
[0020] Among them, β ij represents the large-scale fading between two APs labeled i and j, h ij represents the small-scale fading between two APs labeled i and j;
[0021] Step S103: Based on the consideration of RF link mismatch factors, a bidirectional service channel model between the AP and the UE transceiver is established.
[0022]
[0023] In the formula, and They represent the RF link gain coefficients in the UE receiver and transmitter labeled k respectively; and They represent the RF link gain coefficients in the receiver and transmitter at the AP end, which are labeled m; n m→k and n k→m They respectively represent the additive white Gaussian noise of the UE-end receiver labeled k and the AP-end receiver labeled m in the bidirectional traffic channel.
[0024] Establish a bidirectional internal channel model between any two APs labeled i and j
[0025]
[0026] In the formula, and They represent the RF link gain coefficients in the receiver and transmitter of the AP end labeled i respectively; and They represent the RF link gain coefficients in the AP receiver and transmitter labeled j, respectively; n i→j and n j→i They respectively represent the additive white Gaussian noise of the AP end receiver labeled j and the AP end receiver labeled i in the bidirectional service channel.
[0027] Furthermore, in step S2, in order to solve the channel non-reciprocity problem caused by RF link mismatch, an air interface calibration problem in a non-cellular MIMO scenario is proposed with the goal of minimizing the expected value of the relative error of the channel calibration system, which specifically includes:
[0028] Step S201, considering the air interface partial calibration solution performed on the AP side in a non-cellular large-scale scenario, without the assistance of the user, only the internal channels between APs on the base station side need to be collected;
[0029] Step S202: Channel estimation of uplink receiving end channel state information CSIR performed at the AP side k→m , the channel estimation of downlink CSIR performed on the UE side g m→k ;
[0030] Define the partial channel calibration coefficient c on the AP side m The expression is:
[0031]
[0032] Using this calibration factor c m After calibrating the uplink channel, the following expression can be obtained:
[0033]
[0034] Step S203: Measure the large-scale fading information β of the wireless channel between any two APs ij ,in and storing this information in an offline database;
[0035] Step S204: construct the expectation of the relative error of the channel calibration system as the loss function
[0036]
[0037] In the formula, π β represents the strategy for calculating the channel calibration coefficients, ε i represents the relative error between the calculated channel calibration coefficient and the accurate value; J E (π β ) adopts the expected form of relative error as the loss function of the system;
[0038] Step S205, by optimizing and calculating the channel calibration strategy π β , find the optimal strategy to minimize the loss function;
[0039] When calculating the calibration coefficient of AP labeled B, the calibration coefficient of AP labeled A is used. Then the relative error of the channel calibration coefficient ε B It can be expressed as
[0040]
[0041] In the formula, n A→B and n B→A It represents the mutually independent AWGN in the bidirectional channel, which has zero mean and variance σ 2 Circularly symmetric complex Gaussian distribution;
[0042] Relative error of channel calibration coefficient ε B The expressions of the expectation and second-order moment of are
[0043]
[0044] In the formula, Since the RF link gain coefficients in this expression are all band estimators and obey the independent and identically distributed assumption, the error expectation expression is only related to Directly proportional.
[0045] Furthermore, in step S3, based on the large-scale fading prior information between channels within the AP, the directed tree structure of the graph theory model is used to construct a description of the topological calibration relationship between APs in the non-cellular communication system. Specifically, the following steps are included:
[0046] Step S301: collect all distributed APs in the non-cellular wireless communication system Modeled as a set of nodes in graph theory; the path of each AP calibration coefficient calculation scheme is modeled as a set of directed edges ε between nodes, then the topological relationship in the non-cellular wireless communication system can be expressed as a directed connection graph
[0047] Step S302: Model the reference AP specified in the non-cellular wireless communication system as the root node of the channel calibration tree, and all other AP nodes are leaf nodes of the channel calibration tree;
[0048] Step S303: According to the path connection relationship of each AP calibration coefficient calculation scheme, define the adjacency matrix W = (ω ij ) M×M
[0049]
[0050] In the formula, the connection and disconnection relationship of the edge can be mathematically expressed as ω ij =1 and ω ij =0;
[0051] Step S304: Establish a spanning tree strategy using the current spanning tree strategy according to the topological structure of the channel calibration tree. tree The weighted least squares error loss function under
[0052]
[0053] In the formula, A vector of estimated values of the channel calibration coefficients.
[0054] Furthermore, due to the calibration strategy π tree A directed tree structure is used, and the error of the weighted least squares error loss function can be taken to 0. Specifically, the tree graph has the following properties
[0055]
[0056] Therefore, the method of calculating the calibration coefficient is also simplified, and there is no need to solve overdetermined equations.
[0057] Furthermore, based on the modeling method of the channel calibration tree, the channel calibration tree graph structure constructed in step S301 and step S302 needs to satisfy the following three constraints as a directed tree model:
[0058] (1) The in-degree of the root node in the channel calibration tree is 0;
[0059] (2) The in-degree of a leaf node in the channel calibration tree is 1;
[0060] (3) In the channel calibration tree, there is reachability from the root node to any leaf node.
[0061] Furthermore, based on the calibration tree structure constraint and the calibration coefficient calculation method based on the calibration tree, a calibration path set is set. is an alternating sequence of nodes and edges experienced in the path, serving as a channel calibration process from the reference AP to each of the other target APs; according to the relative error proportional relationship obtained in step S205, a relative error expected value set of the channel calibration coefficient is set where δ i The inverse of the large-scale fading between APs along the calibration path β -1 replace.
[0062] Furthermore, step S4 specifically includes the following steps:
[0063] Step S401: Input the distributed AP set Large-scale fading information between any two APs
[0064] Step S402: Initialize the calibration path to Initialize the expected value of the relative error of the channel calibration coefficient Where v1 represents the node corresponding to the AP numbered 1 in the channel calibration tree, e 1i Indicates the directed edge between the AP labeled 1 and the AP labeled i in the channel calibration tree. i Indicates the node corresponding to the AP labeled i in the channel calibration tree.
[0065] Step S403: The values are sorted from large to small, and the sorting results are represented by indexes as (s1, s2, ..., s M-1 ), assign the value of variable m to 1;
[0066] Step S404: Set the value of variable i to s m , find the path Γ in the channel calibration tree graph structure j e ji v i , node v j is node v i The parent node of , and solve the following equation
[0067]
[0068] Traverse except v i For all nodes other than itself, the parent node number corresponding to the minimum relative error expectation is stored in the variable k;
[0069] Step S405: Determine Is it established? If so, perform the following operations:
[0070] (1) Replace the original edge ε←ε\{e ji}∪{e ki};
[0071] (2) Update the calibration path set Γ according to the tree graph ε tree and the expected set of relative errors of the channel calibration coefficients Δ;
[0072] (3) Update the sort index (s1, s2, ..., s M-1 );
[0073] (4) Assign the value of variable m to 0.
[0074] Step S406, add 1 to the value of m;
[0075] Step S407, iteratively updating the channel calibration tree structure for multiple times, and continuously repeating the calibration path replacement operation of steps S404 to S406 until m≥M;
[0076] Step S408: calibrate the tree ε according to the obtained channel tree Calculate the adjacency matrix W tree , which can be expressed in physical sense as the weight in the weighted least squares error loss function. For W tree The weights satisfy ω ij = 1 for all nodes v i and node v j , let's assume that v i Yes j The parent node of the parent node, based on the channel calibration relationship between the parent node and the child node and the two-way channel estimation result, calculates the calibration coefficient as follows
[0077]
[0078] Thus, the calibration coefficients of each AP based on the channel calibration tree are obtained.
[0079] The beneficial effects of the present invention are:
[0080] 1. Compared with the traditional cellular wireless communication scenario, the present invention uses the step S1, considers the 6G non-cellular wireless communication model based on distributed AP which is different from the traditional wireless communication scenario, and has higher spectrum efficiency than the traditional communication system;
[0081] 2. Compared with the traditional communication scenario, the present invention uses steps S2 to S4, taking into account the impact of RF link mismatch on the reciprocity of the bidirectional channel; facing the non-reciprocity problem under the characteristic of distributed AP, a channel calibration relationship framework is established based on the topological structure of the directed tree, and a 6G non-cellular massive MIMO channel calibration method based on the spanning tree algorithm is proposed;
[0082] 3. Compared with the traditional channel calibration method, the present invention uses the step S4. The channel method based on the spanning tree algorithm has higher accuracy and the lowest computational complexity compared with the traditional channel calibration algorithm Argos, traditional LS, and grouped LS algorithms. BRIEF DESCRIPTION OF THE DRAWINGS
[0083] Figure 1 It is a flowchart of a 6G non-cellular massive MIMO channel calibration method based on a spanning tree algorithm provided in Embodiment 1 of the present invention;
[0084] Figure 2 is a schematic diagram of a non-cellular wireless communication framework provided in Embodiment 1 of the present invention;
[0085] Figure 3 It is a flow chart for establishing the channel calibration problem model provided in Example 1 of the present invention. DETAILED DESCRIPTION
[0086] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0087] A 6G non-cellular massive MIMO channel calibration method based on spanning tree algorithm, such as Figure 1 As shown, the following steps are included:
[0088] Step S1, constructing a 6G non-cellular massive MIMO wireless communication system in time division duplex (TDD) mode, establishing a two-way service channel model between AP and UE and a two-way internal channel model between any two APs, both of which are collectively referred to as a two-way wireless channel model, and the channel distortion factors considered include: large-scale fading (path loss, shadow effect), small-scale fading, and additive Gaussian white noise;
[0089] The construction of a 6G non-cellular massive MIMO wireless communication system in a time division duplex (TDD) mode, establishing a bidirectional service channel model between an AP and a UE and a bidirectional internal channel model between any two APs, specifically includes:
[0090] Step S101: Set the distribution area of the non-cellular communication scenario, set the number of distributed APs and UEs and their distribution locations, and establish the communication model as follows: Figure 2 As shown, specifically including:
[0091] In the Cartesian coordinate system, M APs are set with uniformly distributed two-dimensional coordinate positions in the non-cellular communication scene area, and the set of all APs is defined as
[0092] In the Cartesian coordinate system, K UEs are set with uniformly distributed two-dimensional coordinate positions in the non-cellular communication scenario area, and the set composed of all UEs is defined as
[0093] Step S102: Considering the large-scale fading and small-scale fading of the communication channel, the bidirectional service physical channel between the AP and the UE is modeled as:
[0094]
[0095] In the formula, β mk Indicates the label is The AP is labeled Large-scale fading between UEs, h mk Indicates the label is The AP is labeled Similarly, the bidirectional internal physical channel between any two APs labeled i and j is modeled as
[0096]
[0097] Among them, β ij represents the large-scale fading between two APs labeled i and j, h ij represents the small-scale fading between two APs labeled i and j;
[0098] Step S103: Based on the consideration of RF link mismatch factors, a bidirectional service channel model between the AP and the UE transceiver is established.
[0099]
[0100] In the formula, and They represent the RF link gain coefficients in the UE receiver and transmitter labeled k respectively; and They represent the RF link gain coefficients in the receiver and transmitter at the AP end, which are labeled m; n m→k and nk→m They respectively represent the additive white Gaussian noise of the UE-end receiver labeled k and the AP-end receiver labeled m in the bidirectional traffic channel.
[0101] Establish a bidirectional internal channel model between any two APs labeled i and j
[0102]
[0103] In the formula, and They represent the RF link gain coefficients in the receiver and transmitter of the AP end labeled i respectively; and They represent the RF link gain coefficients in the AP receiver and transmitter labeled j, respectively; n i→j and n j→i They respectively represent the additive white Gaussian noise of the AP end receiver labeled j and the AP end receiver labeled i in the bidirectional service channel.
[0104] Step S2: A complete bidirectional channel includes not only the physical channel, but also the RF link, digital-to-analog / analog-to-digital (DA / AD) converter, filter, mixer and antenna of the transceiver. The introduction of these hardware devices causes the entire wireless channel to be non-reciprocal. Aiming at the channel non-reciprocity problem caused by RF link mismatch, the air interface channel calibration problem in a non-cellular massive MIMO scenario is proposed with the goal of minimizing the expected value of the relative error of the channel calibration system.
[0105] Based on the bidirectional service channel model of the RF link mismatch factor, it can be seen that due to the different channel gain coefficients generated during hardware manufacturing, the uplink and downlink channels no longer have channel reciprocity, which can be expressed as:
[0106]
[0107] In step S2, in order to solve the channel non-reciprocity problem caused by RF link mismatch, the air interface calibration problem in a non-cellular MIMO scenario is proposed with the goal of minimizing the expected value of the relative error of the channel calibration system. The model establishment process is as follows: Figure 3 As shown, specifically including:
[0108] Step S201, considering the air interface partial calibration solution performed on the AP side in a non-cellular large-scale scenario, without the assistance of the user, only the internal channels between APs on the base station side need to be collected;
[0109] Step S202: Channel estimation of uplink receiving end channel state information CSIR performed at the AP side k→m , the channel estimation of downlink CSIR performed on the UE side g m→k ;
[0110] Define the partial channel calibration coefficient c on the AP side m The expression is:
[0111]
[0112] Using this calibration factor c m After calibrating the uplink channel, the following expression can be obtained:
[0113]
[0114] From the formula, it can be seen that after the uplink channel is calibrated, there is only a difference in the multiplication coefficient related to the UE-side RF link between the uplink channel and the downlink channel. This multiplication coefficient will not affect the beam pattern generated by the AP's downlink transmission, and therefore will not affect the performance of the communication system.
[0115] Step S203: Measure the large-scale fading information β of the wireless channel between any two APs ij ,in and storing this information in an offline database;
[0116] Step S204: construct the expectation of the relative error of the channel calibration system as the loss function
[0117]
[0118] In the formula, π β represents the strategy for calculating the channel calibration coefficients, ε i represents the relative error between the calculated channel calibration coefficient and the accurate value. In actual non-cellular wireless communication systems, due to the influence of external factors such as noise, the accurate value of the channel calibration coefficient cannot be obtained. Therefore, J E (π β ) adopts the expected form of relative error as the loss function of the system.
[0119] Step 205, by optimizing and calculating the channel calibration strategy π β , find the optimal strategy to minimize the loss function;
[0120] For example, when calculating the calibration coefficient of AP labeled B, the calibration coefficient of AP labeled A is used. Then the relative error of the channel calibration coefficient is ε B It can be expressed as
[0121]
[0122] In the formula, n A→B and n B→A It represents the mutually independent AWGN in the bidirectional channel, which has zero mean and variance σ 2Circularly symmetric complex Gaussian distribution.
[0123] Relative error of channel calibration coefficient ε B The expressions of the expectation and second-order moment of are
[0124]
[0125] In the formula, Since the RF link gain coefficients in this expression are all band estimators and obey the independent and identically distributed assumption, the error expectation expression is only related to Directly proportional.
[0126] Step S3, according to the optimization goal described in step S2, a channel calibration relationship framework is constructed, based on the large-scale fading prior information between channels within the AP, a directed tree structure of a graph theory model is used to construct a description of the topological calibration relationship between APs in the non-cellular communication system: the reference AP specified by the system is used as the root node of the channel calibration tree, other APs are leaf nodes of the channel calibration tree, and the directed edges of the channel calibration tree represent the calibration calculation relationship between APs;
[0127] The specific steps include:
[0128] Step S301: collect all distributed APs in the non-cellular wireless communication system Modeled as a set of nodes in graph theory; the path of each AP calibration coefficient calculation scheme is modeled as a set of directed edges ε between nodes, then the topological relationship in the non-cellular wireless communication system can be expressed as a directed connection graph
[0129] Step S302: Model the reference AP specified in the non-cellular wireless communication system as the root node of the channel calibration tree, and all other AP nodes are leaf nodes of the channel calibration tree;
[0130] Step S303: According to the path connection relationship of each AP calibration coefficient calculation scheme, define the adjacency matrix W = (ω ij ) M×M
[0131]
[0132] In the formula, the connection and disconnection relationship of the edge can be mathematically expressed as ω ij =1 and ω ij =0;
[0133] Step S304: Establish a spanning tree strategy using the current spanning tree strategy according to the topological structure of the channel calibration tree. tree The weighted least squares error loss function under
[0134]
[0135] In the formula, A vector of estimated values of the channel calibration coefficients.
[0136] Furthermore, due to the calibration strategy π tree A directed tree structure is used, and the error of the weighted least squares error loss function can be taken to 0. Specifically, the tree graph has the following properties
[0137]
[0138] Therefore, the method of calculating the calibration coefficient is also simplified, and there is no need to solve overdetermined equations.
[0139] Furthermore, based on the modeling method of the channel calibration tree, the channel calibration tree graph structure constructed in step S301 and step S302 needs to satisfy the following three constraints as a directed tree model:
[0140] (1) The in-degree of the root node in the channel calibration tree is 0;
[0141] (2) The in-degree of a leaf node in the channel calibration tree is 1;
[0142] (3) In the channel calibration tree, there is reachability from the root node to any leaf node.
[0143] Furthermore, based on the calibration tree structure constraint and the calibration coefficient calculation method based on the calibration tree, a calibration path set is set. is an alternating sequence of nodes and edges experienced in the path, which serves as the channel calibration process from the reference AP to other target APs; according to the relative error proportional relationship obtained in step 205, the relative error expected value set of the channel calibration coefficient is set where δ i The inverse of the large-scale fading between APs along the calibration path β -1 replace.
[0144] Step S4, based on the directed tree model constructed in step S3, a spanning tree algorithm is proposed to minimize the expected value of the system relative error; in the proposed algorithm, the traditional Argos calibration topology structure is used as the initialization of the channel calibration tree, and the topological relationship of the channel calibration is optimized by modifying the directed connection relationship between the nodes in the channel calibration tree, and the optimal channel calibration tree is obtained by continuous iteration and calculation, and the channel calibration coefficient of each AP in the non-cellular system is calculated based on the channel calibration tree.
[0145] The specific steps include:
[0146] Step S401: Input the distributed AP set Large-scale fading information between any two APs
[0147] Step S402: Initialize the calibration path to Initialize the expected value of the relative error of the channel calibration coefficient Where v1 represents the node corresponding to the AP numbered 1 in the channel calibration tree, e 1i Indicates the directed edge between the AP labeled 1 and the AP labeled i in the channel calibration tree. i Indicates the node corresponding to the AP labeled i in the channel calibration tree.
[0148] Step S403: The values are sorted from large to small, and the sorting results are represented by indexes as (s1, s2, ..., s M-1 ), assign the value of variable m to 1;
[0149] Step S404: Set the value of variable i to s m , find the path Γ in the channel calibration tree graph structure j e ji v i , node v j is node v i The parent node of , and solve the following equation
[0150]
[0151] Traverse except v i For all nodes other than itself, the parent node number corresponding to the minimum relative error expectation is stored in the variable k;
[0152] Step S405: Determine Is it established? If so, perform the following operations:
[0153] (1) Replace the original edge ε←ε\{e ji}∪{e ki};
[0154] (2) Update the calibration path set Γ according to the tree graph ε tree and the expected set of relative errors of the channel calibration coefficients Δ;
[0155] (3) Update the sort index (s1, s2, ..., s M-1 );
[0156] (4) Assign the value of variable m to 0.
[0157] Step S406, add 1 to the value of m;
[0158] Step S407, iteratively updating the channel calibration tree structure for multiple times, and continuously repeating the calibration path replacement operation of steps S404 to S406 until m≥M;
[0159] Step S408: calibrate the tree ε according to the obtained channel tree Calculate the adjacency matrix W tree , which can be expressed in physical sense as the weight in the weighted least squares error loss function. For W tree The weights satisfy ω ij = 1 for all nodes v i and node v j , let's assume that v i Yes j The parent node of the parent node, based on the channel calibration relationship between the parent node and the child node and the two-way channel estimation result, calculates the calibration coefficient as follows
[0160]
[0161] Thus, the calibration coefficients of each AP based on the channel calibration tree are obtained.
[0162] Specifically, in this embodiment, the channel calibration method based on the spanning tree algorithm is compared with other traditional algorithms such as Argos, least squares (LS), and grouped least squares (Grouped LS) algorithms in terms of system calibration performance under different transmitter transmit powers. The horizontal axis represents the transmit power of the transmitter, and the vertical axis represents the mean square error between the obtained channel calibration coefficient and the standard value. As the transmitter power increases, the mean square error of the calibration decreases; comparing the spanning tree algorithm we proposed with other algorithms, our algorithm has better performance under the condition of the same transmit power. This shows that the air interface channel calibration method proposed in this embodiment effectively solves the calibration problem under the non-cellular communication system, and improves the channel calibration accuracy with low computational complexity.
Claims
1. A 6G non-cellular massive MIMO channel calibration method based on a spanning tree algorithm, characterized in that: The steps include: Step S1, constructing a 6G non-cellular massive MIMO wireless communication system in time division duplex (TDD) mode, establishing a two-way service channel model between AP and UE and a two-way internal channel model between any two APs, both of which are collectively referred to as a two-way wireless channel model, and the channel distortion factors considered include large-scale fading, small-scale fading and additive white Gaussian noise; Step S2: A complete bidirectional channel includes not only the physical channel but also the RF link of the transceiver. The introduction of the RF link of the transceiver causes the entire wireless channel to be non-reciprocal. Aiming at the channel non-reciprocity problem caused by RF link mismatch, an air interface channel calibration problem in a non-cellular massive MIMO scenario is proposed with the goal of minimizing the expected value of the relative error of the channel calibration system. Step S3: According to the air interface channel calibration problem described in step S2, a channel calibration relationship framework is constructed. Based on the large-scale fading prior information between channels within the AP, a directed tree structure of a graph theory model is used to construct a description of the topological calibration relationship between APs in the non-cellular communication system: the reference AP specified by the system is used as the root node of the channel calibration tree, and other APs are leaf nodes of the channel calibration tree. The directed edges of the channel calibration tree represent the calibration calculation relationship between APs. Step S4, a spanning tree algorithm is proposed to minimize the expected value of the system relative error; in the proposed algorithm, the traditional Argos calibration topology structure is used as the initialization of the channel calibration tree, and the topological relationship of the channel calibration is optimized by modifying the directed connection relationship between the nodes in the channel calibration tree, and the optimal channel calibration tree is obtained by continuous iteration and calculation, and the channel calibration coefficient of each AP in the non-cellular system is calculated based on the channel calibration tree; The 6G non-cellular massive MIMO wireless communication system in the time division duplex (TDD) mode is constructed in step 1, and a bidirectional wireless channel model between the AP and the UE and between any two APs is established, specifically including: Step S101: Setting a distribution area of a non-cellular communication scenario, setting the number of distributed APs and UEs and their distribution locations, and establishing a communication model, specifically including: In the Cartesian coordinate system, M APs are set with uniformly distributed two-dimensional coordinate positions in the non-cellular communication scene area, and the set of all APs is defined as In the Cartesian coordinate system, K UEs are set with uniformly distributed two-dimensional coordinate positions in the non-cellular communication scenario area, and the set composed of all UEs is defined as Step S102: Considering the large-scale fading and small-scale fading of the communication channel, the bidirectional service physical channel between the AP and the UE is modeled as: In the formula, β mk Indicates the label is The AP is labeled Large-scale fading between UEs, h mk Indicates the label is The AP is labeled Similarly, the bidirectional internal physical channel between any two APs labeled i and j is modeled as Among them, β ij represents the large-scale fading between two APs labeled i and j, h ij represents the small-scale fading between two APs labeled i and j; Step S103: Based on the consideration of RF link mismatch factors, a bidirectional service channel model between the AP and the UE transceiver is established. In the formula, and They represent the RF link gain coefficients in the UE receiver and transmitter labeled k respectively; and They represent the RF link gain coefficients in the receiver and transmitter at the AP end, which are labeled m; n m→k and n k→m They respectively represent the additive white Gaussian noise of the UE receiver labeled k and the AP receiver labeled m in the bidirectional traffic channel; Establish a bidirectional internal channel model between any two APs labeled i and j In the formula, and They represent the RF link gain coefficients in the receiver and transmitter at the AP end labeled i respectively; and They represent the RF link gain coefficients in the AP receiver and transmitter labeled j, respectively; n i→j and n j→i denote the additive white Gaussian noise of the AP end receiver labeled j and the AP end receiver labeled i in the bidirectional service channel respectively; In step S2, in order to solve the channel non-reciprocity problem caused by RF link mismatch, an air interface calibration problem in a non-cellular MIMO scenario is proposed with the goal of minimizing the expected value of the relative error of the channel calibration system, which specifically includes: Step S201, considering the air interface partial calibration solution performed on the AP side in a non-cellular large-scale scenario, without the assistance of the user, only the internal channels between APs on the base station side need to be collected; Step S202: Channel estimation of uplink receiving end channel state information CSIR performed at the AP side k→m , the channel estimation of downlink CSIR performed on the UE side g m→k ; Define the partial channel calibration coefficient c on the AP side m The expression is: Using this calibration factor c m After calibrating the uplink channel, the following expression can be obtained: Step S203: Measure the large-scale fading information β of the wireless channel between any two APs ij ,in and storing this information in an offline database; Step S204: construct the expectation of the relative error of the channel calibration system as the loss function In the formula, π β represents the strategy for calculating the channel calibration coefficients, ε i represents the relative error between the calculated channel calibration coefficient and the accurate value; J E (π β ) adopts the expected form of relative error as the loss function of the system; Step S205, by optimizing and calculating the channel calibration strategy π β , find the optimal strategy to minimize the loss function; When calculating the calibration coefficient of AP labeled B, the calibration coefficient of AP labeled A is used. Then the relative error of the channel calibration coefficient ε B It can be expressed as In the formula, n A→B and n B→A It represents the mutually independent AWGN in the bidirectional channel, which has zero mean and variance σ 2 Circularly symmetric complex Gaussian distribution; Relative error of channel calibration coefficient ε B The expressions of the expectation and second-order moment of are In the formula, Since the RF link gain coefficients in this expression are all band estimators and are subject to the independent and identically distributed assumption, the error expectation expression is only related to is proportional to; In step S3, based on the large-scale fading prior information between channels within the APs, a directed tree structure of a graph theory model is used to construct a topological calibration relationship between APs in the non-cellular communication system, which specifically includes the following steps: Step S301: collect all distributed APs in the non-cellular wireless communication system Modeled as a set of nodes in graph theory; the path of each AP calibration coefficient calculation scheme is modeled as a set of directed edges ε between nodes, then the topological relationship in the non-cellular wireless communication system can be expressed as a directed connection graph Step S302: Model the reference AP specified in the non-cellular wireless communication system as the root node of the channel calibration tree, and all other AP nodes are leaf nodes of the channel calibration tree; Step S303: According to the path connection relationship of each AP calibration coefficient calculation scheme, define the adjacency matrix W = (ω ij ) M×M In the formula, the connection and disconnection relationship of the edge can be mathematically expressed as ω ij =1 and ω ij =0; Step S304: Establish a spanning tree strategy using the current spanning tree strategy according to the topological structure of the channel calibration tree. tree The weighted least squares error loss function under In the formula, A vector of estimated values of channel calibration coefficients; Step S4 specifically includes the following steps: Step S401: Input the distributed AP set Large-scale fading information between any two APs Step S402: Initialize the calibration path to Initialize the expected value of the relative error of the channel calibration coefficient Where v1 represents the node corresponding to the AP numbered 1 in the channel calibration tree, e 1i Indicates the directed edge between the AP labeled 1 and the AP labeled i in the channel calibration tree. i Indicates the node corresponding to the AP labeled i in the channel calibration tree; Step S403: The values are sorted from large to small, and the sorting results are represented by indexes as (s1, s2, ..., s M-1 ), assign the value of variable m to 1; Step S404: Set the value of variable i to s m , find the path Γ in the channel calibration tree graph structure j e ji v i , node v j is node v i The parent node of , and solve the following equation Traverse except v i For all nodes other than itself, the parent node number corresponding to the minimum relative error expectation is stored in the variable k; Step S405: Determine Is it established? If so, perform the following operations: (1) Replace the original edge ε←ε\{e ji }∪{e ki }; (2) Update the calibration path set Γ according to the tree graph ε tree and the expected set of relative errors of the channel calibration coefficients Δ; (3) Update the sort index (s1, s2, ..., s M-1 ); (4) Assign the value of variable m to 0; Step S406, add 1 to the value of m; Step S407, iteratively updating the channel calibration tree structure for multiple times, and continuously repeating the calibration path replacement operation of steps S404 to S406 until m≥M; Step S408: calibrate the tree ε according to the obtained channel tree Calculate the adjacency matrix W tree , which can be expressed in physical sense as the weight in the weighted least squares error loss function. For W tree The weights satisfy ω ij = 1 for all nodes v i and node v j , let's assume that v i Yes j The parent node of the parent node, based on the channel calibration relationship between the parent node and the child node and the two-way channel estimation result, calculates the calibration coefficient as follows Thus, the calibration coefficients of each AP based on the channel calibration tree are obtained.
2. According to the 6G non-cellular massive MIMO channel calibration method based on the spanning tree algorithm according to claim 1, it is characterized in that: Since the calibration strategy π tree A directed tree structure is used, and the error of the weighted least squares error loss function can be taken to 0. Specifically, the tree graph has the following properties 3. According to the 6G non-cellular massive MIMO channel calibration method based on the spanning tree algorithm according to claim 2, it is characterized in that: Based on the modeling method of the channel calibration tree, the channel calibration tree graph structure constructed in step S301 and step S302 needs to satisfy the following three constraints as a directed tree model: (1) The in-degree of the root node in the channel calibration tree is 0; (2) The in-degree of a leaf node in the channel calibration tree is 1; (3) In the channel calibration tree, there is reachability from the root node to any leaf node.
4. According to claim 3, a 6G non-cellular massive MIMO channel calibration method based on a spanning tree algorithm is characterized in that: Based on the calibration tree structure constraints and the calibration tree-based calibration coefficient calculation method, the calibration path set is set is an alternating sequence of nodes and edges experienced in the path, serving as a channel calibration process from the reference AP to each other target AP; according to the relative error proportional relationship obtained in step S205, a relative error expected value set of the channel calibration coefficient is set where δ i The inverse of the large-scale fading between APs along the calibration path β -1 replace.
Citation Information
Patent Citations
MIMO (Multiple Input Multiple Output) system reciprocity calibration method in time division duplex mode
CN108667534A
Improved Gaussian tree approximate message transfer detection algorithm in large-scale MIMO system
CN109981151A
User-assisted method for calibrating channel of multi-antenna small base station for time division duplexing
WO2017166432A1
