Smart reflector assisted multi-cell coordinated distributed communication system design method

Abstract

The application relates to the technical field of wireless communication, in particular to a smart reflecting surface assisted multi-cell cooperation distributed communication system design method. The technical scheme of the application is that a message passing algorithm based on Bayesian estimation is used for precoding processing of a transmitting end and a smart reflecting surface, so that the bottleneck of low coverage capacity and low edge user spectrum efficiency of an existing multi-cell cellular network is broken through.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and specifically to a design method for a distributed communication system with intelligent reflector-assisted multi-cell cooperation. Background Technology

[0002] In the future 6G era, very large-scale antenna systems (VMS) have a wide range of applications, but they also face a series of challenges, including severe path loss at high frequencies, limited communication distance and coverage, high difficulty in channel measurement and modeling, large signal processing computations, and heavy fronthaul pressure. For macro base station deployments, strengthening hotspot coverage and filling blind spots will become standard practice for future 6G antenna systems. In reality, the actual deployment of VMS is not optimistic. While stacking VMS improves coverage, the resulting increased signal processing complexity and cost limit the actual deployment capabilities of the equipment.

[0003] Smart reflectors, as a breakthrough energy-saving technology for future 6G, can control the propagation environment by introducing additional scattering. Their controlled scattering characteristics can generate passive beamforming for the desired receiver, achieving high beamforming gain and suppressing co-channel interference, potentially revolutionizing wireless communication. Ideally, smart reflectors can create an intelligent, programmable, and controllable wireless propagation environment, bringing new degrees of freedom to wireless network optimization beyond traditional transceiver designs. The combination of passive and active precoding is key to improving network capacity using smart reflectors. Unlike traditional precoding performed only at the base station, joint precoding in smart reflector-based wireless systems involves the joint design of beamforming vectors and smart reflector element phase shifts at the base station. Summary of the Invention

[0004] The purpose of this invention is to propose a distributed communication system design method suitable for multi-cell cooperation assisted by intelligent reflectors. The communication system designed using this distributed method exhibits low computational complexity and scalability. The technical solution of this invention overcomes the bottlenecks of low coverage capacity and low spectral efficiency for edge users in existing multi-cell cellular networks by precoding the base station and intelligent reflector using a message passing algorithm based on Bayesian estimation.

[0005] Consider as Figure 1 The downlink massive MIMO system based on a reconfigurable smart reflector shown is assumed to have L multi-antenna base stations and N antennas for the l-th base station. l And N l >1, the number of multi-antenna users is K, and the number of antennas for the k-th user is J. k The number of received data streams is M k Furthermore, J k >M k>1, the r-th intelligent reflective surface includes G r Each metasurface unit, and G r >1; Define the user's received signal data stream codebook as This represents the codebook of the known signal data stream received by the k-th multi-antenna user. Assume the virtual received signal model is as follows:

[0006]

[0007] in, This represents the signal received by the known k-th multi-antenna user. This represents the baseband channel matrix from all multi-antenna base stations in multiple cells within a cellular system to multiple smart reflectors; This represents the baseband channel matrix from multiple smart reflectors to multiple multi-antenna users; This represents the baseband channel matrix from multiple multi-antenna base stations to multiple multi-antenna users; Let x = Fy be the transmitted signal vector x = Fy derived from the joint precoding of all base stations based on the known user-side signal y, where F is the precoding matrix on the base station side, and Φ is the precoded transmit signal vector of the first base station; r-r =diag(Θ1, ...,Θ) R Let be the joint relative angle matrix of all intelligent reflective surfaces derived from the known user-side codebook y, where Let be the passive adjustment relative angle matrix of the r-th intelligent reflective surface; The additive white Gaussian noise matrix has a variance of σ. 2 I M .

[0008] By combining multiple multi-antenna base stations, multiple multi-antenna users, and multiple smart reflectors from multiple cells, a distributed communication system is formed. This system consists of L multi-antenna base stations as transmitters, K multi-antenna users as receivers, and R smart reflectors as reflectors. Within the cellular system, the channels between multiple base stations and multiple users in multiple cells are not fully linked, but rather partially linked. That is, a user only receives signals from its associated base stations; for example, the k-th user only receives signals from the set of base stations. The signal; a base station only provides signals to its associated users, for example, the first base station only provides signals from the user set. Transmit a signal.

[0009] In the cellular system, the channels between multiple base stations in multiple cells and multiple smart reflectors are not fully linked, but rather partially linked. That is, a smart reflector only reflects signals from its associated base station. For example, the r-th smart reflector only receives signals from the set of base stations. The signal; a base station only provides a signal to its associated metasurface, for example, the first base station only provides a signal from the set of smart reflective surfaces. Transmit a signal.

[0010] In the cellular system, the channels between multiple smart reflectors in multiple cells and multiple users are not fully linked, but rather partially linked. That is, a user only receives reflected signals from its associated metasurface; for example, the k-th user only receives signals from the set of base stations. The signal; a hypersurface only provides signals to its associated users, for example, the r-th hypersurface only provides signals from the user set. Transmit a signal.

[0011] The signal vector actually received by the user can be written as Assumption

[0012]

[0013] The main objective is to [do something] within the known [context]. Under the conditions, from Obtaining the Φ r-r The estimate, and then the calculation From observation data Solving for an unknown matrix Φ r-r This involves a bilinear estimation problem. To solve this problem using probabilistic estimation (Bayesian estimation), the problem form is transformed into:

[0014]

[0015] Specifically, in order to make the received data stream signal infinitely close to Transmitting signals to base stations and the intelligent reflective surface reflects the signal Φ r-r To perform precoding, auxiliary variables are first introduced to establish a system model equivalent to the received signal model:

[0016]

[0017] b = Φ r-r a = diag(Θ)a (Formula 5)

[0018]

[0019] Among them, matrices a and b are auxiliary variable matrices introduced to establish an equivalent system model. The structural features of this equivalent model are the key to the design of a low-complexity receiver algorithm. The (4th) and (6th) sub-models are linear models of x, and the (5th) sub-model is Φ. r-rThe pointwise bilinear model, combined with the equivalent model above, considers the estimation problem at the receiver from the perspective of Bayesian statistical inference. First, the conditional probabilities are obtained from the equivalent model above, as follows:

[0020]

[0021]

[0022]

[0023] The first conditional probability is Based on the equivalent formula (1), we obtain that, under the condition that b and x are known, the received quantity The conditional probability of the first term; the conditional probability of the second term is... According to the equivalent model formula (2), it represents the known and Under the condition of auxiliary variables The conditional probability of the third term is... According to the equivalent model formula (2), the auxiliary variable represents the condition where x is known. The conditional probability; furthermore, in the above formula, δ(·) represents the cyclic symmetric complex Gaussian probability density function, and δ(·) represents the delta function. The subscript kx represents the k-th column of column vector y. j The subscripts of the elements and other variables have the same meaning.

[0024] At this time Φ r-r The joint posterior probability of x and x under a given received signal is:

[0025]

[0026] in It is a uniform distribution of reflection angles [0, 2π] on the intelligent reflective surface. The initial value is a Gaussian distribution. Statistical inference requires obtaining Φ based on this joint posterior probability. r-r The marginal posterior probability estimate of x is used as its estimate, but the marginal posterior probability estimate involves high-dimensional integral complexity, which is too high to implement. Therefore, in this embodiment, a novel distributed message passing algorithm is designed to approximate the statistical inference problem of the joint posterior probability description.

[0027] Specifically, such as Figure 2 The factor graph model shown is a graphical representation of the joint posterior probability. Rectangular factor nodes represent conditional probabilities in the joint posterior probability, and circular variable nodes represent random variables. Each edge is treated as a unit to discuss the message update method. (The symbols are notation.) and These represent the messages from variable node i to factor node j and from factor node j to variable node i, respectively, during the t-th iteration. For example... Figure 4 As shown, the specific message passing formula is expressed as follows:

[0028] S1: The message passing algorithm at the receiving end is an iterative algorithm. The algorithm first performs initialization:

[0029]

[0030]

[0031] In the above formula, This represents the signal estimate on the nth transmit antenna of the first base station in this iteration, initially set to 0; This represents the estimated offset angle on the g-th reflecting unit of the r-th smart reflector in this iteration. It is initially set to 0, and all other variables are intermediate values. Their purpose is to ultimately assist the base station and the metasurface signal; they have no actual physical meaning. Furthermore, in the above formula… The Frobenius norm of a matrix. Indicates the r-th g The probability that a reflection unit takes the i-th angle from the i-th discrete offset angle set is initially set to... I represents the total number of possible discrete offset angles.

[0032] S2: Update module Input to module The message variance and mean are as follows:

[0033]

[0034]

[0035] in, As an intermediate variable, Estimate the mean squared error for this variable. and Represents what was obtained in the previous iteration and

[0036] S3: Update module Output about variables Message variance, mean: message (where subscript) For factor nodes variance (abbreviation) with the mean as follows:

[0037]

[0038]

[0039] in, and All of these are intermediate variables.

[0040] S4: Combining modules Output about The message and its mean and variance are calculated: all inflow variable nodes The mean and variance of the message set are as follows:

[0041]

[0042]

[0043] in, Auxiliary variables in the equivalent system model The estimated mean, This corresponds to the estimated mean square error;

[0044] in Used to obtain and in the formula This represents a value among the discrete quantized values ​​of the reflection angle.

[0045] S5: Combining modules Output about The message and its mean and variance are calculated: all inflow variable nodes The mean and variance of the message set are as follows:

[0046]

[0047]

[0048] in Auxiliary variables in the equivalent system model The estimated mean, To estimate the mean square error, See S4 for the updated formula.

[0049] S6: Update Module Output about The message, input it Additional information obtained:

[0050]

[0051] That is, from the factor node To variable node The news, then The input is:

[0052] Combination Input and After quantization, the available angle set is used, and corresponding angle quantization operations are performed, further updating the external information:

[0053]

[0054] in, represent The probability of; This will be used in subsequent iterations.

[0055] S7: Combined Module and Output about The message updates its mean and variance, and all incoming variable nodes. The mean and variance of the message set are as follows:

[0056]

[0057]

[0058] Let g be the estimated signal value on the g-th reflecting unit within the r-th intelligent reflecting surface segment. To estimate the mean square error, the estimation of the intelligent reflector signal in this algorithm iteration is now complete.

[0059] S8: Update input module About The probability of taking each discrete quantized value, probability as follows:

[0060]

[0061] Among them, for Note that when hour

[0062] S9: Update Module Output about The variance and mean of the message; the message The variance and mean are as follows:

[0063]

[0064]

[0065] in, and For estimation Intermediate quantities in the service.

[0066] S10: Update Module Output about Message variance, mean, and message update The variance and mean are as follows:

[0067]

[0068]

[0069] in and For estimation Intermediate quantities in the service.

[0070] S11: Module Output about Message Input It also outputs external information, as well as information about S9 and S10. Update the variable value. The input is:

[0071] The external information used in this input / output is:

[0072]

[0073] S12: Combining modules and Output about Message and update its mean and variance, update and All inflow variable nodes The mean and variance of the message set are as follows:

[0074]

[0075]

[0076] in, This represents the signal estimate on the nth transmit antenna of the first base station in this iteration. This represents the corresponding mean square error of the estimation, thus completing the base station signal estimation in this algorithm iteration.

[0077] S13: Update module Output about message variance, mean, message The variance and mean are as follows:

[0078]

[0079]

[0080] in and Represents what was obtained in the previous iteration and Furthermore, this step is a condition for forming an iterative closed loop.

[0081] S14: If the algorithm converges, i.e., in the two iterations... If the percentage change is less than one-thousandth, output the result of the last iteration. and As and If the estimated value does not converge, return to S2 to continue the iteration.

[0082] It should be noted that in the specific expression of the message passing formula, the updates of most variables are only used as intermediate results to form an iterative closed loop, and have no actual physical meaning. This allows each smart reflector unit to transmit information using its phase changes at the smart reflector end, thereby improving the information transmission rate. Since the smart reflector information is modulated on the base station incident signal, no additional radio frequency components are needed. Furthermore, the reliability of transmission is ensured by introducing modulation and coding techniques in the smart reflector and base station. On the other hand, at the receiving end, a novel message passing algorithm is designed under the Bayesian framework. The proposed algorithm can utilize the modulation and coding information of the base station and the smart reflector end. Each iteration characterizes the estimated signal value and the mean square error of the base station and the smart reflector end. In the iteration, the estimated value of the previous iteration is used as a soft pilot to obtain a more accurate estimate until convergence. The algorithm complexity is linearly related to the number of iterations T and quadratically related to the total number of base station antennas N, the total number of smart reflector units G, and the total number of receiving antennas M. This process is implemented in parallel using matrix multiplication and vector multiplication to achieve computational correlation. Here, the transmitting end refers to multiple multi-antenna base stations in multiple cells within a cellular system, and the receiving end refers to multiple multi-antenna users in multiple cells within a cellular system. Attached Figure Description

[0083] Figure 1 This is a downlink transmission block diagram of the distributed precoding device for a cellular network based on multi-cell cooperation proposed in this example;

[0084] Figure 2 The factor graph corresponding to the distributed message passing algorithm for the multiple multi-antenna users proposed in this example;

[0085] Figure 3 This is a two-dimensional geographic location distribution map of the communication system proposed in this example;

[0086] Figure 4The flowchart shows the distributed message passing algorithm used for multiple multi-antenna users proposed in this example.

[0087] Figure 5 This is a graph showing the cumulative distribution function of spectral efficiency for the three central cell users in this example. Detailed Implementation

[0088] The present invention will now be described in detail with reference to the accompanying drawings, so that those skilled in the art can better understand the present invention.

[0089] This embodiment illustrates the precoding and modulation process based on a specific implementation case. The specific parameter settings can be configured as follows, considering a smart reflector-assisted multi-user MIMO communication system, such as... Figure 3 As shown, this is a cellular network consisting of 12 units. Figure 3 As shown, each cell has 3 multi-antenna base stations (25m high), 27 multi-antenna users (165m high), and 9 smart reflectors (10m high). The vertex 'a' of each hexagonal cell is 125 meters. The three base stations are positioned on the three bisectors of the cellular network, so that their centers coincide with the cell center. All smart reflectors and users are deployed on a ring centered on each cell ([0.12a-0.97a]).

[0090] Assuming that both the direct channel from the base station to the user / smart reflector and the channel from the smart reflector to the user follow Rayleigh fading, i.e. in and ρ l,k and This represents the calculation of large-scale coefficients for a massive MIMO system, including array gain and link path loss determined by the three-dimensional positions of base station users and smart reflectors. Furthermore, This only represents the path loss of the direct connection between the smart reflector r and the user k, which is determined by the three-dimensional positions of the smart reflector and the user, as shown in the figures. Figure 3 As shown.

[0091] like Figure 3 As shown, based on the above parameter settings, the specific steps of this simulation are as follows:

[0092] consider Figure 3 The entire cellular system comprises 36 base stations, 108 smart reflectors, and 324 users. Each base station is equipped with 32 transmit antennas, each smart reflector has 32 reflective elements, and each user has 4 receive antennas, i.e., L=36, R=108, K=324, N1=32, G r =32, J k=4, corresponding to the received signal model. My algorithm for optimizing the transmission symbol x and phase angle is quantized using the value Ms=4, that is...

[0093] pass Figure 5 It is understood that, compared with the "passive beamforming" curve, this embodiment overcomes the bottleneck of the small capacity of existing wireless communication cellular networks. Compared with the alternating optimization algorithm curve, the user algorithm of this embodiment has superior spectral efficiency performance. The "lower bound of user spectral efficiency (single connection)" curve is the curve of this embodiment when a single user receives signals from only one base station and one smart reflector, serving as the lower bound of user spectral efficiency performance. The "upper bound of user spectral efficiency (full connection)" curve is the curve of this embodiment when a single user receives signals from all base stations and all smart reflectors in the entire network, serving as the upper bound of user spectral efficiency performance. It can be seen that this embodiment overcomes the bottlenecks of coverage capacity and low spectral efficiency of edge users in existing multi-cell cellular networks.

Claims

1. A design method for a distributed communication system with intelligent reflective surface-assisted multi-cell cooperation, characterized in that, By combining multiple multi-antenna base stations, multiple multi-antenna users, and multiple smart reflectors from multiple cells, a multi-cell cooperative distributed communication system is formed. This system consists of L multi-antenna base stations as transmitters, K multi-antenna users as receivers, and R smart reflectors as reflectors. The l-th base station has N antennas. l And N l >1, the number of antennas for the k-th user is J k The number of user streams received is M k Furthermore, J k >M k >1, the r-th intelligent reflective surface includes G r Each metasurface unit, and G r >1; Define the received signal data stream codebook at the receiving end as , The codebook representing the received signal data stream of the known k-th multi-antenna user is used to establish the received signal model as follows: ， in, , This represents the signal received by the known k-th multi-antenna user. This represents the baseband channel matrix from the base station to the smart reflector; This represents the baseband channel matrix from the smart reflector to the user; This represents the baseband channel matrix from the base station to the user; For known received signals The base station jointly precoded transmitted signal vector is derived from the reverse. ,in Let be the precoding matrix of the base station, and Let N be the precoded transmit signal vector of the l-th multi-antenna base station. l It is the number of antennas in the l-th multi-antenna base station; For known received signals The joint relative angle matrix of all intelligent reflective surfaces is derived by inversion, where Let r be the passive adjustment relative angle matrix of the r-th intelligent reflective surface. Let be the offset angle of the g-th transmitting unit of the r-th smart reflector; The matrix is ​​an additive white Gaussian noise matrix; In the multi-cell cooperative distributed communication system, the channel from the base station to the user is defined as a partial link, meaning a user only receives the transmitted signal from the base station associated with it, and a base station only provides the signal to the user associated with it; the signal from the base station to the smart reflector is also a partial link, meaning a smart reflector only reflects the signal from the base station associated with it, and a base station only provides the signal to the smart reflector associated with it; the signal from the smart reflector to the user is also a partial link, meaning a user only receives the reflected signal from the smart reflector associated with it, and a smart reflector only provides the signal to the user associated with it; a message passing algorithm based on Bayesian estimation is used to precode the base station and the smart reflector, specifically as follows: Introducing auxiliary variables to establish a system model equivalent to the received signal model: ， ， ， Among them, matrix and This is the matrix of auxiliary variables introduced; using the Bayesian estimation method, the conditional probability obtained from the equivalent system model is: ， ， ， in, Represents known and Under these conditions, the amount of data received The conditional probability; Represents known and Under the condition of auxiliary variables The conditional probability; Represents known Under the condition of auxiliary variables The conditional probability; This represents a cyclically symmetric complex Gaussian probability density function. Represents the delta function. Subscript Represents column vectors The The meaning of the subscripts for other variables is similar; get and The joint posterior probability given the received signal is: ， in, Denotes a generalized normalization factor constant. It is a smart reflective surface with a high reflection angle. Uniform distribution on The initial value is a Gaussian distribution; based on this joint posterior probability, we obtain... and The marginal posterior probability estimate is used as its estimated value. Specifically, a message-passing algorithm is used to approximate the joint posterior probability. The specific solution process is as follows: S1. Initialization: ， ， ， in, This represents the signal estimate on the nth transmit antenna of the l-th base station in this iteration, initially set to 0; This represents the estimated offset angle on the g-th reflecting unit of the r-th smart reflective surface in this iteration, initially set to 0; the other variables are intermediate quantities with no actual physical meaning. The Frobenius norm of a matrix. Indicates the first The probability that a reflection unit takes the i-th angle from the i-th discrete offset angle set is initially set to... , This represents the total number of possible discrete offset angles. S2, Update Module , Input to module The message variance and mean are as follows: ， ， in, As an intermediate variable, Estimate the mean squared error for this variable. and Represents what was obtained in the previous iteration and ; S3, Update Module Output about variables message variance, mean, message variance with the mean as follows: ， ， in, and All are intermediate variables; S4, Combined Module , Output about The message is processed and its mean and variance are calculated for all incoming variable nodes. The mean and variance of the message set are as follows: ， ， in, Auxiliary variables in the equivalent system model The estimated mean, This corresponds to the estimated mean square error; ,in Used to obtain and , in the formula This represents a discrete quantized value for the reflection angle. S5, Combined Module , Output about The message is processed and its mean and variance are calculated for all incoming variable nodes. The mean and variance of the message set are as follows: ， ， in Auxiliary variables in the equivalent system model The estimated mean, To estimate the mean square error, See S4 for the updated formula; S6, Update Module Output about The message, input it Unexpected information: ； That is, from the factor node To variable node The news, then The input is: ; combination Input and After quantization, the available angle set is determined, and corresponding angle quantization operations are performed, followed by further updates to external information. ， in, represent The probability of; This will be used in subsequent iterations; S7, Combined Module and Output about The message updates its mean and variance, and all incoming variable nodes. The mean and variance of the message set are as follows: ， ； This represents the signal estimate on the g-th reflecting unit within the r-th intelligent reflecting surface segment. To estimate the mean square error, the estimation of the intelligent reflective surface signal in this algorithm iteration is now complete. S8, Update Input Module About The probability of taking each discrete quantized value, probability as follows: ， Among them, for , until hour ; S9, Update Module Output about The variance and mean of the message; the message The variance and mean are as follows: ， ， in, and For estimation Intermediate quantities in the service; S10, Update Module Output about Message variance, mean, and message update The variance and mean are as follows: ， ， in and For estimation Intermediate quantities in the service; S11, Module , Output about Message input It also outputs external information, as well as information about S9 and S10. Update the variable value. The input is: The external information used in this input / output is: ； S12, Combined Module , and Output about Message and update its mean and variance, update and All inflow variable nodes The mean and variance of the message set are as follows: ， ， in, This represents the signal estimate on the nth transmit antenna of the l-th base station in this iteration. This represents the corresponding mean square error of the estimation, thus completing the base station signal estimation in this algorithm iteration; S13, Update Module Output about message variance, mean, message The variance and mean are as follows: ， ， in and Represents what was obtained in the previous iteration and ; S14. If the algorithm converges, that is, if the two iterations are successful... , If the percentage change is less than one-thousandth, output the result of the last iteration. and As and If the estimated value does not converge, return to S2 to continue the iteration.

Images