A terahertz cell-free system null interference suppression method and device

Abstract

This invention discloses a method and device for suppressing null interference in a terahertz cellular-free system, comprising: acquiring an offline feature dictionary, acquiring the grid index of the target user in the current time slot and the grid index of the interfering user that poses strong interference to the target user, and jointly encoding them with the access point index serving the target user to generate a spatial state hash fingerprint; querying a cache library using the spatial state hash fingerprint as the key; if the cache query is successful, directly reading the corresponding beamforming weight vector; if the cache query is unsuccessful, extracting the array steering vector of the grid where the target user and the interfering user are located according to the offline feature dictionary, constructing the beamforming weight vector of the target user, and storing it in the cache library; weighting the received signal according to the beamforming weight vector of the target user, and outputting the signal after interference suppression. This invention can balance null robustness and real-time performance in dynamic and uncertain environments.

Description

Technical Field

[0001] This invention relates to terahertz interference suppression technology, and more particularly to a method and apparatus for suppressing zero-dip interference in terahertz non-cellular systems. Background Technology

[0002] Terahertz (THz) bands possess abundant spectrum resources and are a core candidate band for 6G; however, their extremely high path loss necessitates the use of massive MIMO arrays to generate extremely narrow directional beams for compensation. To further improve coverage performance, cellular-free massive MIMO systems distribute a large number of access points (APs) and coordinate their control through a central processing unit to eliminate cell boundary effects and achieve macro diversity gain. Combining terahertz technology with a cellular-free architecture can effectively address the challenges of high-frequency band attenuation, providing ultra-high-speed communication services for high-density indoor scenarios.

[0003] While the combination of terahertz technology and non-cellular multi-access point (MAP) architecture can significantly improve network capacity, it faces complex multi-AP co-channel interference coupling problems in dense deployments. On the one hand, the extremely narrow beams of terahertz technology are highly sensitive to spatial disturbances; even slight user displacement or positioning deviation can cause beam nulls to miss the target, leading to severe interference leakage. On the other hand, in terahertz non-cellular systems, the extremely short channel coherence time and the large antenna array size mean that obtaining high-precision global CSI not only faces extremely high pilot overhead but is also highly susceptible to severe mismatch in precoding weights due to channel aging or estimation errors. Existing methods generally rely excessively on these rapidly changing instantaneous channel characteristics, greatly limiting the transmission reliability and throughput of the system in dynamic environments. Furthermore, existing methods generally lack effective mechanisms for reusing historical experience, resulting in a large amount of redundant computational overhead. Therefore, there is an urgent need for a low-latency interference suppression method that can actively sense the spatial topology of interference and balance null robustness and real-time performance in dynamic and uncertain environments. Summary of the Invention

[0004] To address the problems existing in the prior art, the purpose of this invention is to provide a method and device for suppressing zero-traps interference in terahertz non-cellular systems that can actively sense the spatial topology of interference and balance zero-traps robustness and real-time performance in dynamic and uncertain environments.

[0005] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0006] A method for suppressing zero-dip interference in a terahertz cellular-free system includes:

[0007] Obtain an offline feature dictionary, wherein the offline feature dictionary includes a two-dimensional Sinc tapered matrix and an array guidance tensor dictionary. The array guidance tensor dictionary is formed by the array guidance vectors of each grid formed after each access point is discretized to the user's plane. The two-dimensional Sinc tapered matrix is ​​determined by a preset zero-depression widening control parameter.

[0008] Obtain the grid index of the target user in the current time slot and the interfering user that causes strong interference to the target user, and jointly encode it with the access point index of the target user to generate a spatial state hash fingerprint.

[0009] The cache library is queried using the spatial state hash fingerprint as the key; if the cache query is successful, the corresponding beamforming weight vector is read directly, wherein the cache library is used to store the beamforming weight vectors of the users contained in the spatial state hash fingerprint;

[0010] If the cache query fails, the array steering vectors of the grids where the target user and the interfering user are located are extracted according to the offline feature dictionary. A simulated precoding matrix is ​​constructed, and the interference plus noise covariance matrix is ​​reduced to the radio frequency subspace corresponding to the simulated precoding matrix through algebraic equivalence tapering operation. The baseband weight vector is solved on the reduced interference plus noise covariance matrix. The beamforming weight vector of the target user is formed by the baseband weight vector and the simulated precoding matrix and stored in the cache library.

[0011] The received signal is weighted according to the beamforming weight vector of the target user, and the signal after interference suppression is output.

[0012] Furthermore, the offline feature dictionary is obtained through offline calculation, and the specific calculation method includes:

[0013] In a terahertz non-cellular indoor scene, the plane where the user is located is discretized into several uniform grids.

[0014] Calculate the array steering vector from all antenna elements at each access point to each grid.

[0015] A two-dimensional Sinc tapered matrix is ​​generated based on preset zero-depression widening control parameters.

[0016] The array guide vectors from all access points to all grids are used to form an array guide tensor dictionary, which is then combined with the two-dimensional Sinc tapered matrix as an offline feature dictionary.

[0017] Furthermore, the method for generating the spatial state hash fingerprint includes:

[0018] The access point estimates the location of the target user based on the uplink pilot signal, and then finds the index of the nearest grid based on the estimated location, which is used as the grid index of the target user.

[0019] Identify the set of interfering users that pose a strong interference to the target user, wherein the interfering users are those whose relative power weight to the target user is greater than a preset threshold;

[0020] The access point index of the target user, the grid index of the target user, the grid index of each user in the set of interfering users, and the radio frequency link status are jointly encoded to form a spatial state hash fingerprint. The radio frequency link status is used to characterize the arrangement of array steering vectors corresponding to each radio frequency link on the access point of the target user in the current time slot.

[0021] Furthermore, the method for identifying the interfering user specifically includes:

[0022] For other users who do not belong to the same access point service as the target user, the relative power weight is determined based on the difference between their macro-scale gain and the target user's macro-scale gain.

[0023] Users whose relative power weight is greater than a preset threshold are identified as interference users of the target user, forming a set of interference users of the target user.

[0024] Furthermore, when the number of key-value pair entries in the cache exceeds the set capacity limit, the least recently used eviction policy is executed.

[0025] Furthermore, the method for constructing the simulated precoding matrix includes:

[0026] Construct an empty analog precoding matrix;

[0027] Extract the array steering vectors of all users belonging to the same access point service as the target user from the offline feature dictionary, and assign each array steering vector as a column to the analog precoding matrix;

[0028] Extract the array guide vector of each user's neighboring grid from the offline feature dictionary and assign it to the remaining unassigned columns of the simulated precoding matrix until all columns have been assigned.

[0029] Furthermore, the step of reducing the dimension of the interference plus noise covariance matrix to the radio frequency subspace corresponding to the analog precoding matrix through algebraic equivalence cone reduction specifically includes:

[0030] For each interfering user of the target user, the baseband dimension-reduced covariance component of the interfering user is determined based on the array steering vector of the interfering user grid, the analog precoding matrix, and the two-dimensional Sinc tapered matrix.

[0031] The baseband covariance components of the interfering user are weighted and fused to obtain the total baseband interference, which is then combined with the noise to obtain the dimension-reduced interference plus noise covariance matrix.

[0032] Furthermore, the step of solving for the baseband weight vector on the dimensionality-reduced interference plus noise covariance matrix specifically includes:

[0033] The baseband weight vector is obtained by multiplying the inverse matrix of the dimensionality-reduced interference plus noise covariance matrix, the conjugate transpose of the analog precoding matrix, and the array steering vector of the target user grid.

[0034] Furthermore, the beamforming weight vector is normalized according to the target user's transmit power and then used as a weight to weight the received signal.

[0035] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described above.

[0036] Compared with the prior art, the beneficial effects of this invention are:

[0037] 1. This invention establishes an interference statistical model with spatial broadening characteristics by tapering the covariance matrix, and completes digital beam optimization by forming a radio frequency subspace with hybrid beams. This enables the system to robustly suppress the dominant interference direction and its spatial neighborhood even in the presence of positioning errors and user micro-movements, thereby significantly improving the interference suppression capability and spatial robustness of terahertz non-cellular networks.

[0038] 2. This invention meshes the physical space and designs a "state hash fingerprint" and state-driven caching architecture based on relative network topology. By sensing slowly changing spatial topological features to assist beam generation, the system can reuse historical interference countermeasure experience; when the cache hits, the final hybrid beamforming weight vector is directly output, thereby reducing the dependence on instantaneous small-scale CSI, reducing the overhead of repeated covariance assembly and weight solving, and improving the system's online response latency. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the terahertz zero-dip interference suppression method for cellular-free systems provided in this embodiment of the invention.

[0040] Figure 2 This is a simulation result comparison of the system and rate of the method provided in this embodiment of the invention with the traditional method under different movement errors.

[0041] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0043] Figure 1 A method for suppressing zero-dip interference in terahertz cellular-free systems is provided, such as... Figure 1 and Figure 2 As shown, the process includes the following steps S101 to S105.

[0044] S101. Obtain the offline feature dictionary.

[0045] The offline feature dictionary includes a two-dimensional Sinc tapered matrix and an array-guided tensor dictionary. The array-guided tensor dictionary is formed by the array-guided vectors of each grid formed by discretizing each access point to the user's plane. The two-dimensional Sinc tapered matrix is ​​determined by preset zero-depression widening control parameters.

[0046] The offline feature dictionary is obtained through offline pre-computation. The pre-computation steps specifically include:

[0047] S1011 Discretize the plane where the user is located in the terahertz non-cellular indoor scene into several uniform grids.

[0048] In this embodiment, the terahertz non-cellular system consists of One access point and It consists of single-antenna users, and each access point is equipped with a total of 1,000 antennas. A uniform planar antenna array with [number] antenna elements, and configured... One radio frequency link, and These represent the number of antenna elements in the horizontal and vertical directions of the uniform planar antenna array, respectively.

[0049] Deploy a terahertz non-cellular system with a length of [length in the x-direction]. The length in the y direction is The length in the z-direction is In indoor scenarios, single-antenna users are distributed at a height of [missing information]. The horizontal plane, the access point is installed in the indoor scene at a height of The location.

[0050] With preset step size For heights of in the x and y directions respectively The user plane is uniformly discretized and sampled to generate a set of grid coordinates in the x-direction. The set of grid coordinates in the y direction The total number of grid points is , , Index all grid cells according to their coordinates to obtain the grid index. , Indicates the grid index in the x-direction. This represents the grid index in the y-direction.

[0051] S1012. Calculate the array steering vector from all antenna elements at each access point to each grid.

[0052] For each access point and each grid Corresponding three-dimensional coordinates , express Calculate the array steering vector from access point b to the grid based on the corresponding coordinates. The array steering vector is calculated using a near-field spherical wave model, with each access point equipped with a model containing... A uniform planar antenna array with n elements, where the m-th element in the horizontal direction and the n-th element in the vertical direction are defined. To grid The precise distance is:

[0053]

[0054]

[0055] In the formula, Represents the array reference element to the grid. straight-line distance This indicates the spacing between array elements in the horizontal direction. This indicates the spacing between array elements in the vertical direction. Represents a grid The coordinate difference in the horizontal direction relative to the reference array element. Represents a grid The coordinate difference in the vertical direction between the relative array reference element and the reference element, which can be any one of the antenna elements.

[0056] The guiding vector elements are:

[0057]

[0058] In the formula, This indicates the operating wavelength of the terahertz carrier wave.

[0059] The array steering vector is: .

[0060] S1013. Generate a two-dimensional Sinc tapered matrix according to the preset zero-depression broadening control parameters.

[0061] To introduce spatial angular perturbations in the direction of the interference source to achieve a wide null, the construction of the two-dimensional Sinc tapering matrix specifically includes:

[0062] For a uniform planar antenna array, construct one-dimensional Sinc tapered filtering matrices in the horizontal and vertical directions respectively. and Its corresponding matrix elements , They respectively satisfy:

[0063]

[0064]

[0065] In the formula, , These are the set horizontal and vertical zero-depression widening control parameters, respectively.

[0066] Through Kronek product Synthesize a globally shared two-dimensional Sinc tapered matrix .

[0067] S1014. Form an array guidance tensor dictionary from the array guidance vectors of all access points to all grids, and use it together with the two-dimensional Sinc tapered matrix as an offline feature dictionary.

[0068] The set of array steering vectors from all access points to all grids is stored as a four-dimensional tensor, forming an array steering tensor dictionary. ,in Mapped to Middle elements The dictionary is pre-calculated offline once during system deployment, resides in memory during runtime, and is shared and accessible across the entire network.

[0069] The two-dimensional Sinc tapered matrix and array steering vector step complete offline pre-calculation; the interference covariance is spatially broadened and modeled by the two-dimensional Sinc tapered matrix, thereby expanding the original interference statistical model that is concentrated in a single angular direction into a tapered covariance model that covers a certain spatial angular domain, providing a statistical basis for constructing a spatial wide null in the subsequent digital beamforming stage.

[0070] S102. Obtain the grid index of the target user in the current time slot and the interfering user that causes strong interference to the target user, and jointly encode it with the access point index of the target user to generate a spatial state hash fingerprint.

[0071] Step S102 specifically includes:

[0072] S1021. The access point estimates the location of the target user based on the uplink pilot and finds the index of the nearest grid based on the estimated location, which serves as the corresponding user grid index.

[0073] In this embodiment of the terahertz non-cellular system, a non-coherent joint transmission mode is used to provide services to users on the same time-frequency resources. The system's communication channel adopts a terahertz three-dimensional spatial channel model that considers molecular absorption, dynamic blocking, and multipath fading, and introduces equivalent noise caused by transceiver hardware impairment and molecular absorption in the solution.

[0074] During the current time slot of the online phase, the access point can obtain the target user via uplink pilot. Estimated position coordinates Search and The most recent grid, use that grid index as the grid index for user k. At the same time, for the entire network When multiple users perform the same operation, a global grid index cache set is created. This avoids duplicate searches.

[0075] S1022. Identify the set of interfering users that pose a strong interference to each user.

[0076] Interfering users are defined as those whose relative power weight to the current user is greater than a preset threshold. The specific method for identifying interfering users is as follows:

[0077] For each user not belonging to the same access point service as the target user k, the relative power weight is determined based on the difference between their macroscopic large-scale gain and the target user's macroscopic large-scale gain. Specifically, for the ... Let there be *n* access points, and let the set of users they currently serve be *x*. For all non-service users across the entire network Calculate its relative power weight with respect to the target user k. The calculation formula is:

[0078]

[0079] In the formula, For the macroscopic, large-scale gain from non-service user j to access point b; The macroscopic large-scale gain from user k to access point b is expressed in decibels.

[0080] relative power weight Users identified as interference users of target user k are sorted in descending order of relative power weight and form a set of interference users of target user k. , Indicates the preset threshold;

[0081] Interference user set The system can record the target user's k-grid index, the effective interfering user's grid index, and the corresponding relative power weights. .

[0082] S1023. Jointly encode the access point index of the target user, the grid index of the target user, the grid index of each user in the set of interfering users, and the radio frequency link status to form a spatial state hash fingerprint.

[0083] Specifically, the access point index of the target user k is... The user grid index and the set of disturbance states RF link status Perform joint hash encoding as a spatial state hash fingerprint This will be used as the unique identifier for subsequent cached queries.

[0084] RF link status The array steering vector arrangement, used to characterize the access point of the target user in the current time slot, corresponds to the arrangement of array steering vectors in each column of the analog precoding matrix. Each column in the analog precoding matrix corresponds to one radio frequency chain, and each column contains one array steering vector. The radio frequency link state, as a component of the spatial state hash fingerprint f, enables the identification of whether the current spatial topology configuration duplicates historical states, thereby achieving beam weight cache reuse.

[0085] S103. Query the cache library using the spatial state hash fingerprint as the key; if the cache query is successful, directly read the corresponding beamforming weight vector.

[0086] The cache library uses a key-value pair storage structure with access order records, storing several key-value pairs. The key is the user's spatial state hash fingerprint, and the value is the beamforming weight vector of the user contained in the spatial state hash fingerprint. The specific calculation method for the beamforming weight vector is described in step S104. Initially, the cache library is empty. Only after the beamforming weight vector is calculated in step S104 and stored in the cache library will key-value pairs containing the user's spatial state hash fingerprint and the beamforming weight vector of the user contained in the spatial state hash fingerprint exist.

[0087] When a cache lookup using the spatial state hash fingerprint as the key is successful, the system does not need to perform complex matrix operations; it can directly read the final beamforming weight vector corresponding to the state hash fingerprint f from the cache. The system outputs the result; simultaneously, it updates the access order record of the state hash fingerprint f in the cache to maintain the priority of this active state. This mechanism replaces the traditional high-dimensional matrix inversion calculation with a super-fast table lookup, greatly reducing the response latency of online beamforming.

[0088] When a cache lookup miss occurs, it indicates that the system is currently facing a completely new spatial interference topology. At this point, the system jumps to execute the complete solution process of step S104 to recalculate the baseband weight vector and the analog precoding matrix.

[0089] In addition, as a preferred cache management mechanism, to prevent the risk of memory overflow caused by long-term system operation, when the number of entries in the cache library exceeds the preset capacity limit, the system automatically executes the least recently used eviction policy to actively pop the oldest expired entries, thereby strictly controlling the cache size within the preset capacity limit and ensuring the long-term stability of the system's memory usage.

[0090] The caching mechanism is used to reuse historical beam results under the same or equivalent spatial topology to reduce the overhead of repeated covariance assembly and weight solving.

[0091] S104. If the cache query fails, extract the array steering vectors of the grids where the target user and the interfering user are located according to the offline feature dictionary, construct a simulated precoding matrix, and reduce the dimension of the interference plus noise covariance matrix to the radio frequency subspace corresponding to the simulated precoding matrix through algebraic equivalence tapering operation. Solve for the baseband weight vector on the reduced interference plus noise covariance matrix, and form the beamforming weight vector of the target user according to the baseband weight vector and the simulated precoding matrix, and write it into the cache library.

[0092] This step specifically includes:

[0093] S1041. Based on the offline feature dictionary, extract the array steering vector corresponding to the user's grid to form the main beam column of the analog precoding matrix, and extract the array steering vector of the user's neighboring grids to form the neighborhood auxiliary beam column in the analog precoding matrix.

[0094] Methods for constructing analog precoding matrices include:

[0095] S10411, Construct a empty analog precoding matrix .

[0096] S10412. For each user who belongs to the same access point b as the target user k... According to its grid index And the access point index b that serves it, extract the corresponding array steering vector from the offline feature dictionary. The values ​​are then assigned to the analog precoding matrix in the order of the service user list of access point b. The former Each column represents the main beam column for the corresponding user. For example, if user k is the kth user in the service user list, then its array steering vector is located in the analog precoding matrix. The kth column;

[0097] S10413, For each user The array steering vectors of the adjacent grids are extracted sequentially and assigned to the analog precoding matrix. In the remaining unassigned columns, until all columns have been assigned values, the remaining columns, excluding the main beam column, form the neighborhood auxiliary beam column. Specifically, the array steering vector of one neighboring grid for each user can be extracted sequentially from the service user list of access point b, and then assigned a value. If the simulated precoding matrix... If the assignment is not yet complete, another round of extraction and assignment of the guide vectors of the neighboring grid array will be performed in sequence until all columns are assigned.

[0098] In this design, the adjacent auxiliary beamline does not directly point towards interfering users. Instead, it constructs multiple spatially correlated steering vectors around the neighboring grid of the serving user. These vectors, together with the main beamline, form an analog precoding matrix, serving as the RF subspace required for digital baseband optimization. The digital baseband then performs weight optimization after covariance matrix taper within this RF subspace, thereby obtaining a larger digital baseband optimization subspace under finite RF chain conditions. This also avoids hardware noise radiation problems caused by the analog end directly facing the interference direction.

[0099] S1042. For each interfering user of the target user, determine the baseband dimension-reduced covariance component of the interfering user based on the array steering vector of the grid where the interfering user is located, the analog precoding matrix, and the two-dimensional Sinc tapering matrix.

[0100] Specifically, for each interfering user of the target user k Utilize its grid Corresponding array steering vector With the analog precoding matrix Constructing a tapered radio frequency matrix ,in For column-by-column and element-by-element Hadamard product, express The conjugate matrix; combined with the two-dimensional Sinc cone matrix, the interfering user is obtained. Baseband dimension reduction covariance components :

[0101]

[0102] In the formula, It is a two-dimensional Sinc tapered matrix. This represents the conjugate transpose of the matrix. By using a two-dimensional Sinc tapered matrix to spatially broaden the covariance components, the original interference statistical model, which was concentrated in a single angular direction, is extended to a tapered covariance model covering a certain spatial angular domain. This provides a statistical basis for constructing spatially wide nulls in the subsequent digital beamforming stage.

[0103] Through the above algebraic equivalence, the online solution dimension is reduced from the antenna dimension to the RF chain dimension, and the explicit construction and inversion of the complete antenna dimension are avoided to prevent interference with the covariance matrix.

[0104] S1043. Weighted fusion of the baseband dimension-reduced covariance components of all interfering users is performed to obtain the total baseband interference, which is then combined with the noise to obtain the dimension-reduced interference-noise covariance matrix. :

[0105]

[0106] In the formula, Indicates the noise variance. This represents the relative power weight.

[0107] S1044. Multiply the inverse matrix of the reduced interference plus noise covariance matrix, the conjugate transpose of the analog precoding matrix, and the array steering vector of the target user grid to obtain the baseband weight vector. ;

[0108]

[0109] S1045. Form beamforming weight vectors based on baseband weight vectors and analog precoding matrices. , beamforming weight vector The value corresponding to the spatial state fingerprint obtained using the target user k number is written into the cache.

[0110] The output of step S1042 is the covariance component of the radio frequency chain dimension, thereby avoiding the direct construction of a large-scale covariance matrix in the antenna dimension. Steps S1043 to S1045 are completed in the radio frequency chain dimension, instead of explicitly generating a complete interference covariance matrix in the antenna dimension.

[0111] S105. Based on the beamforming weight vector of the target user, the received signal is weighted and the signal after interference suppression is output.

[0112] In this embodiment, equal power distribution is adopted, so access point b is a service user. Allocated transmission power , To determine the total transmit power, first adjust the beamforming weight vector. Normalization processing into normalized weights Then, the received signal X(n) is weighted using normalized weights to obtain the signal after interference suppression. .

[0113] Based on the above, in order to verify the effectiveness and effect of the method of the present invention, this embodiment conducts simulation tests, and the results are as follows. Figure 2 As shown. Under ideal, error-free conditions, both the CMT algorithm and MVDR of this invention can achieve high system and speed performance. However, as the position estimation error gradually increases, global ZF and MVDR rely on real-time accurate channel construction to create single-point narrow nulls, making the beam prone to missing targets and gradually degrading in performance. MMSE has a low inherent reference gain, and DCB relies on hard constraints to construct wide nulls, which excessively occupies the degrees of freedom of RF hardware and significantly reduces the gain of the main beam signal. In contrast, this invention relies on prior information from a discrete spatial grid, without relying on instantaneous small-scale channel matrices. It can quickly generate wide null beams based solely on position priors and large-scale gain information, resulting in smoother performance degradation and better robustness under disturbed scenarios.

[0114] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. The embodiment of the present invention is described above. Figure 1 The implementation of the method provides services. For example... Figure 3 As shown, the device may include: a memory 201 storing a computer-executable program; a processor 202 coupled to the memory 201; the processor 202 calls the computer-executable program stored in the memory 201 to perform the steps in the method described in Embodiment 1.

[0115] Memory 201 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, memory 201 may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard disk drive"). A program / utility having a set (at least one) of program modules may be stored, for example, in memory 201. Such program modules include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The computer-executable program of the program modules typically performs the functions and / or methods described in the embodiments of the present invention.

[0116] The processor 202 executes various functional applications and data processing by running programs stored in the memory 201, such as implementing the method provided in Embodiment 1 of the present invention.

[0117] The code of a computer executable program can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages.

[0118] It should be understood that the embodiments and descriptions above are only the principles, main features and advantages of the present invention. Various changes and modifications can be made to the present invention without departing from the spirit and scope of the invention, and all such changes and modifications fall within the protection scope of the present invention.

Claims

1. A method for suppressing zero-dip interference in a terahertz cellular-free system, characterized in that, include: Obtain an offline feature dictionary, wherein the offline feature dictionary includes a two-dimensional Sinc tapered matrix and an array guidance tensor dictionary. The array guidance tensor dictionary is formed by the array guidance vectors of each grid formed after each access point is discretized to the user's plane. The two-dimensional Sinc tapered matrix is ​​determined by a preset zero-depression widening control parameter. Obtain the grid index of the target user in the current time slot and the interfering user that causes strong interference to the target user, and jointly encode it with the access point index of the target user to generate a spatial state hash fingerprint. The cache library is queried using the spatial state hash fingerprint as the key; if the cache query is successful, the corresponding beamforming weight vector is read directly, wherein the cache library is used to store the beamforming weight vectors of the users contained in the spatial state hash fingerprint; If the cache query fails, the array steering vectors of the grids where the target user and the interfering user are located are extracted according to the offline feature dictionary. A simulated precoding matrix is ​​constructed, and the interference plus noise covariance matrix is ​​reduced to the radio frequency subspace corresponding to the simulated precoding matrix through algebraic equivalence tapering operation. The baseband weight vector is solved on the reduced interference plus noise covariance matrix. The beamforming weight vector of the target user is formed by the baseband weight vector and the simulated precoding matrix and stored in the cache library. The received signal is weighted according to the beamforming weight vector of the target user, and the signal after interference suppression is output.

2. The terahertz cellular-free system null interference suppression method according to claim 1, characterized in that, The offline feature dictionary is obtained through offline calculation. The specific calculation method includes: Discretize the user's plane in a terahertz non-cellular indoor scene into several uniform grids; Calculate the array steering vector from all antenna elements at each access point to each grid. A two-dimensional Sinc tapered matrix is ​​generated based on preset zero-depression broadening control parameters; The array guide vectors from all access points to all grids are used to form an array guide tensor dictionary, which is then combined with the two-dimensional Sinc tapered matrix as an offline feature dictionary.

3. The terahertz cellular-free system null interference suppression method according to claim 1, characterized in that, The method for generating the spatial state hash fingerprint includes: The access point estimates the location of the target user based on the uplink pilot signal, and then finds the index of the nearest grid based on the estimated location, which is used as the grid index of the target user. Identify the set of interfering users that pose a strong interference to the target user, wherein the interfering users are those whose relative power weight to the target user is greater than a preset threshold; The access point index of the target user, the grid index of the target user, the grid index of each user in the set of interfering users, and the radio frequency link status are jointly encoded to form a spatial state hash fingerprint. The radio frequency link status is used to characterize the arrangement of array steering vectors corresponding to each radio frequency link on the access point of the target user in the current time slot.

4. The terahertz cellular-free system null interference suppression method according to claim 3, characterized in that, The method for identifying the interfering user is as follows: For other users who do not belong to the same access point service as the target user, the relative power weight is determined based on the difference between their macro-scale gain and the target user's macro-scale gain. Users whose relative power weight is greater than a preset threshold are identified as interference users targeting the target user, thus forming a set of interference users for the target user.

5. The terahertz cellular-free system null interference suppression method according to claim 1, characterized in that, When the number of key-value pairs in the cache exceeds the set capacity limit, the least recently used eviction policy is executed.

6. The terahertz cellular-free system null interference suppression method according to claim 1, characterized in that, The method for constructing the analog precoding matrix includes: Construct an empty analog precoding matrix; Extract the array steering vectors of all users belonging to the same access point service as the target user from the offline feature dictionary, and assign each array steering vector as a column to the analog precoding matrix; Extract the array guide vector of each user's neighboring grid from the offline feature dictionary and assign it to the remaining unassigned columns of the simulated precoding matrix until all columns have been assigned.

7. The terahertz cellular-free system null interference suppression method according to claim 1, characterized in that, The step of reducing the dimension of the interference plus noise covariance matrix to the radio frequency subspace corresponding to the analog precoding matrix through algebraic equivalence specifically includes: For each interfering user of the target user, the baseband dimension-reduced covariance component of the interfering user is determined based on the array steering vector of the grid where the interfering user is located, the analog precoding matrix, and the two-dimensional Sinc tapered matrix. The baseband covariance components of all interfering users are weighted and fused to obtain the total baseband interference, which is then combined with the noise to obtain the dimension-reduced interference plus noise covariance matrix.

8. The terahertz cellular-free system null interference suppression method according to claim 1, characterized in that, The step of solving for the baseband weight vector on the dimensionality-reduced interference plus noise covariance matrix specifically includes: The baseband weight vector is obtained by multiplying the inverse matrix of the dimensionality-reduced interference plus noise covariance matrix, the conjugate transpose of the analog precoding matrix, and the array steering vector of the target user grid.

9. The terahertz cellular-free system null interference suppression method according to claim 1, characterized in that, The beamforming weight vector is normalized according to the target user's transmit power and then used as a weight to weight the received signal.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the computer program to implement the method as described in any one of claims 1-9.

Images