A sparse sounding and rate maximization beamforming method and related apparatus

Abstract

The present application belongs to the field of wireless communication and intelligent network optimization, and discloses a sparse detection and sum rate maximization beamforming method and related device, through constructing a system model containing electromagnetic twin state matrix and graph Laplace matrix, using the last period state and the current sparse observation data to estimate state innovation and update electromagnetic twin state, and calculating the posterior covariance matrix; then, based on the grid where the user is located, the channel covariance matrix is queried, and the optimal beamforming matrix is obtained by solving the sum rate maximization problem under the constraint of total transmit power; finally, the posterior covariance and the query related weight are combined to select the next round of detection grid with the goal of minimizing the upper bound of sum rate estimation error. This method realizes the simultaneous completion of wireless knowledge update, user channel query and detection position optimization under the sparse detection budget, significantly improves the contribution of limited detection resources to the actual communication utility, and finally achieves efficient closed-loop optimization from sparse detection to sum rate maximization.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication and intelligent network optimization technology, and particularly relates to a sparse detection and sum rate maximizing beamforming method and related apparatus. Background Technology

[0002] In sixth-generation mobile communication systems, massive MIMO and high-frequency transmission have become core technologies, and the performance of multi-user beamforming highly depends on channel knowledge. Wireless channels contain multiple information dimensions with different temporal evolution characteristics: instantaneous channel state information changes rapidly with fast fading, while environmental features such as signal-dominant direction of arrival, spatial covariance, and congestion areas are slowly varying variables, with a rate of change far lower than fast fading. These can be learned across multiple detection cycles and reused in measurement-sparse areas. Based on the temporal scale separation characteristics of channel information, radio maps and channel knowledge maps provide environmental knowledge support for wireless resource management and transmission optimization by storing wireless propagation information for specific scenarios. Electromagnetic twin technology further builds a persistent "update-query-measurement" closed-loop mechanism, continuously preserving the environmental propagation state, updating the state based on physical measurement data, and providing state query capabilities for communication decisions.

[0003] Existing radio map reconstruction schemes infer the channel state of unmeasured locations from channel data at measured locations. While effective map completion and application are possible when a complete and representative dataset is available, continuous system operation requires significant probe resources to re-acquire channel data across the entire service area, making it unsuitable for applications with sparse probe budgets. Furthermore, existing probe location selection mechanisms primarily aim to minimize the global uniform map reconstruction error, which fundamentally differs from the actual utility goals of communication systems. With limited probe resources, probe points far from active users only reduce global map errors, offering little to no improvement in beamforming performance for the corresponding user. Conversely, probe points near areas where propagation states change can significantly improve the accuracy of user channel covariance estimation, yet are rarely prioritized by existing probe strategies. Moreover, current technologies cannot simultaneously achieve persistent wireless knowledge updates, user channel covariance queries, and rate-maximizing probe location optimization under sparse probe budget constraints. They lack closed-loop optimization methods that organically integrate sparse observation data, persistent electromagnetic twin states, and beamforming decisions.

[0004] It is evident that existing technologies cannot achieve the organic integration of sparse observation, persistent electromagnetic twin state, and beamforming decision-making under sparse detection conditions, and thus cannot complete the closed-loop optimization from sparse detection to rate maximization. Summary of the Invention

[0005] This invention provides a sparse detection and sum rate maximization beamforming method and related apparatus. Using this method, sparse observation, persistent electromagnetic twin state and beamforming decision can be organically combined under sparse detection conditions, and closed-loop optimization from sparse detection to sum rate maximization can be completed.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A sparse detection and sum rate maximizing beamforming method includes: An electromagnetic twin system model is constructed, which includes a set of candidate detector grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detector observation model, and a graph Laplace matrix; the graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detector grids. Based on the electromagnetic twin system model, the uplink detection observation data corresponding to the selected detection grid set in the current period is obtained. The state innovation quantity of the current period is estimated by combining the electromagnetic twin state matrix saved in the previous period, and the electromagnetic twin state matrix of the current period is updated. At the same time, the posterior covariance matrix corresponding to the current electromagnetic twin state is calculated. Based on the grid location of each active user, query the updated current period electromagnetic twin state matrix to generate the channel covariance matrix for each active user. Based on the channel covariance matrix of each active user, the downlink sum rate maximization problem is solved under the constraint of total transmit power to obtain the optimal downlink beamforming matrix. By combining the posterior covariance matrix and the query-related weights of each candidate probe grid relative to active users, and with the goal of minimizing the upper bound of the sum rate estimation error, a set of candidate probe grids for the next round is selected from the undetected candidate probe grids. The set of candidate probe grids for the next round is fed back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

[0007] Furthermore, the construction of the electromagnetic twin system model includes: The service area is divided into N candidate detection grids, forming a candidate detection grid set; The angle domain is divided into L directions, forming a set of angle domain directions; Construct an electromagnetic twin state matrix of dimension N×L; wherein, the angular power spectrum of the candidate detector grid is: the element in the nth row of the electromagnetic twin state matrix corresponds to the average received power of the nth candidate detector grid in each direction; Define a probe selection matrix; the probe selection matrix is ​​used to select the row vector corresponding to the currently probed grid from the electromagnetic twin state matrix; An uplink sparse detection observation model is established based on the detector selection matrix; the observed value of the uplink sparse detection observation model is equal to the product of the detector selection matrix and the electromagnetic twin state matrix plus the observation noise. A graph Laplacian matrix is ​​constructed to characterize the spatial power correlation between adjacent candidate detector grids, so that locally detected power changes can propagate to adjacent undetected candidate detector grids.

[0008] Furthermore, the process of estimating the state innovation quantity of the current period by combining the electromagnetic twin state matrix stored in the previous period, updating the electromagnetic twin state matrix of the current period, and simultaneously calculating the posterior covariance matrix corresponding to the current electromagnetic twin state includes: The observation residual matrix is ​​obtained by subtracting the product of the current cycle's uplink detection observation data and the detection selection matrix and the electromagnetic twin state matrix saved in the previous cycle. A convex optimization objective function is constructed, which includes a data fitting term, a spatial smoothing term, and a state preservation term. The data fitting term is used to constrain the consistency between the state innovation quantity and the observation residual, the spatial smoothing term realizes the spatial propagation constraint of power change based on the graph Laplacian matrix, and the state preservation term is used to constrain the overall magnitude of the state innovation quantity. The convex optimization objective function is solved based on the normal equation to obtain the state innovation matrix of the current period. The electromagnetic twin state matrix of the previous period is added to the state innovation matrix and then truncated non-negatively to obtain the electromagnetic twin state matrix of the current period. At the same time, the posterior covariance matrix shared by all angular components is solved. The posterior covariance matrix is ​​used to quantify the estimation uncertainty of the current electromagnetic twin state.

[0009] Further, generating the channel covariance matrix corresponding to each active user includes: Construct a direction dictionary matrix consisting of the normalized array response vectors corresponding to each direction; For each active user, the angular power spectrum row vector corresponding to the candidate detection grid where the active user is located is extracted from the electromagnetic twin state matrix of the current period, and the initial channel covariance matrix is ​​calculated by combining it with the direction dictionary matrix. A diagonal loading process is applied to the initial channel covariance matrix, with the diagonal loading amount being proportional to the trace of the initial channel covariance matrix, to obtain the final user channel covariance matrix; wherein, the user channel covariance matrix is ​​used to characterize the second-order statistical characteristics of the channel at the corresponding candidate probe grid location.

[0010] Furthermore, the solution of the downlink sum and rate maximization problem under the constraint of total transmit power includes: Based on the channel covariance matrix of each active user, the reachable rate expression for each active user is constructed, and the reachable rates of all active users are superimposed to obtain the system and rate objective functions. Calculate the Wirtinger gradients of the system and rate objective functions with respect to the beamforming matrix, and construct the near-end lower bound concave quadratic functions of the system and rate. The iteration step size parameter is determined by backtracking search. The optimal solution is obtained for the concave quadratic function of the near-end lower bound under the constraint of total transmit power. The solution is then projected onto the feasible region of transmit power to obtain the iteratively updated beamforming matrix. Repeat the iterations until the system and rate converge, and obtain the final optimal downlink beamforming matrix.

[0011] Furthermore, the selection of the next round of candidate detector grids from the undetected candidate detector grids includes: Based on the spatial distance between each candidate probe grid and each active user, the query-related weights of each candidate probe grid are calculated using the Gaussian kernel function, and a query weight diagonal matrix is ​​constructed. With minimizing the query weighted posterior uncertainty as the proxy optimization objective, the candidate probe grid with the largest single-point variance reduction is selected sequentially from the undetected candidate probe grids until the number of selected candidate probe grids reaches the preset single-round detection limit, thus obtaining the candidate probe grid set for the next round. Wherein, the agent optimization objective corresponds to the upper bound of the system and rate estimation error; Further, the step of sequentially selecting the candidate probe grid that maximizes the reduction in query weighted uncertainty from the unexplored candidate probe grids includes: Using the matrix inversion lemma, we can calculate the updated posterior covariance matrix corresponding to the addition of a single candidate detector grid to the detector set. Based on the query weight diagonal matrix and the posterior covariance matrix before and after the update, the single-point variance reduction corresponding to the candidate probe grid is calculated.

[0012] A sparse detection and sum rate maximizing beamforming system includes: The model building module is used to construct an electromagnetic twin system model, which includes a set of candidate detector grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detector observation model, and a graph Laplace matrix; the graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detector grids. The state update module is used to obtain the uplink detection observation data corresponding to the selected detection grid set in the current period based on the electromagnetic twin system model, estimate the state innovation amount in the current period by combining the electromagnetic twin state matrix saved in the previous period, update the electromagnetic twin state matrix in the current period, and calculate the posterior covariance matrix corresponding to the current electromagnetic twin state. The covariance query module is used to query the updated current period electromagnetic twin state matrix based on the grid location of each active user, and generate the channel covariance matrix corresponding to each active user. The beam optimization module is used to solve the downlink sum and rate maximization problem under the constraint of total transmit power, based on the channel covariance matrix of each active user, to obtain the optimal downlink beamforming matrix. The detector selection module combines the posterior covariance matrix and the query-related weights of each candidate detector grid relative to active users. With the goal of minimizing the upper bound of the sum rate estimation error, it selects the next round of candidate detector grids from the undetected candidate detector grids and feeds the next round of candidate detector grids back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

[0013] A sparse detection and sum rate maximizing beamforming device includes: Memory, used to store computer programs; A processor is used to implement the above-described sparse detection and rate-maximizing beamforming method when executing the computer program.

[0014] A computer-readable storage medium storing a computer program that, when executed by a processor, is used to implement the above-described sparse detection and rate-maximizing beamforming method.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a sparse detection and sum-rate maximization beamforming method. It constructs a system model including an electromagnetic twin state matrix and a graphical Laplacian matrix. The method estimates state innovation using the previous cycle state and current sparse observation data, updates the electromagnetic twin state, and simultaneously calculates the posterior covariance matrix. Then, based on the user's grid location, it queries the channel covariance matrix and solves the sum-rate maximization problem under total transmit power constraints to obtain the optimal beamforming matrix. Finally, combining the posterior covariance and query correlation weights, it selects the next round of detection grids with the goal of minimizing the upper bound of the sum-rate estimation error. This method integrates sparse observation, persistent electromagnetic twin state, and beamforming decision-making in a closed loop. By capturing spatial correlation through the graphical Laplacian matrix, it focuses detection location selection on the region with the greatest impact on user beamforming performance, rather than blindly pursuing the minimization of global map errors. This method achieves simultaneous wireless knowledge updates, user channel queries, and detection location optimization under a sparse detection budget, significantly improving the contribution of limited detection resources to actual communication utility, and ultimately achieving efficient closed-loop optimization from sparse detection to sum-rate maximization. Attached Figure Description

[0016] Figure 1 is a schematic diagram of the processing chain from sparse detection to electromagnetic twin beamforming in an indoor scene provided by an embodiment of the present invention; Figure 2 is a graph showing the relationship between the average sum rate and the random probe grid ratio provided in the embodiment of the present invention; Figure 3 shows the relationship between the normalized mean square error of the angular power spectrum map provided in the embodiment of the present invention and the ratio of random detection grid. Figure 4 is a graph showing the relationship between the average sum and rate after the query sensing and detection update provided in the embodiment of the present invention and the transmission signal-to-noise ratio. Figure 5 is a comparison diagram of closed loop and rate when using rate correlation or random probe selection according to an embodiment of the present invention; Figure 6 is a comparison of the normalized mean square error of the closed-loop angular power spectrum map when using rate correlation or random probe selection according to the embodiments of the present invention; Figure 7 This is a core flowchart of a sparse detection and sum rate maximizing beamforming method provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of a sparse detection and sum rate maximizing beamforming system provided in an embodiment of the present invention. Detailed Implementation

[0017] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0018] As described in the background section, existing technologies struggle to simultaneously achieve persistent wireless knowledge updates, user channel covariance lookup, and rate-maximizing probe location selection within a sparse probe budget.

[0019] To address the aforementioned issues, this embodiment provides a sparse detection and sum rate maximization beamforming method. This method organically combines sparse observation, persistent electromagnetic twin state, and beamforming decision-making, thereby taking into account wireless knowledge persistence updates, user covariance queries, and sum rate-driven detection selection under a sparse detection budget.

[0020] For example, such as Figure 7 As shown, this embodiment provides a sparse detection and sum rate maximizing beamforming method, including: An electromagnetic twin system model is constructed, which includes a set of candidate detector grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detector observation model, and a graph Laplace matrix; the graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detector grids. Based on the electromagnetic twin system model, the uplink detection observation data corresponding to the selected detection grid set in the current period is obtained. The state innovation quantity of the current period is estimated by combining the electromagnetic twin state matrix saved in the previous period, and the electromagnetic twin state matrix of the current period is updated. At the same time, the posterior covariance matrix corresponding to the current electromagnetic twin state is calculated. Based on the grid location of each active user, query the updated current period electromagnetic twin state matrix to generate the channel covariance matrix for each active user. Based on the channel covariance matrix of each active user, the downlink sum rate maximization problem is solved under the constraint of total transmit power to obtain the optimal downlink beamforming matrix. By combining the posterior covariance matrix and the query-related weights of each candidate probe grid relative to active users, and with the goal of minimizing the upper bound of the sum rate estimation error, a set of candidate probe grids for the next round is selected from the undetected candidate probe grids. The set of candidate probe grids for the next round is fed back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

[0021] The sparse detection and rate-maximizing beamforming method provided in this embodiment will be further explained below with reference to the accompanying drawings: For example, this embodiment provides a sparse detection and sum rate maximization beamforming method. This method is geared towards multi-user downlink beamforming scenarios, and its main steps include: constructing an electromagnetic twin system model and saving the angular power spectrum state that varies with location; based on sparse uplink observations, using graph regularization to estimate the updated angular power spectrum innovation and its posterior uncertainty; querying the updated angular power spectrum at active user locations to obtain the user channel covariance; solving the sum rate maximization problem using a projection successive convex approximation under transmit power constraints; selecting the next batch of detection grids based on the query weighted posterior variance reduction criterion, that is, by saving and updating the angular power spectrum electromagnetic twin state, sparse uplink detection is converted into user channel covariance, and the sum rate maximization under power constraints and the selection of the next round of detection locations are further completed; more specific steps are as follows: Step 1: Construct an electromagnetic twin system model. The physical model of the electromagnetic twin system includes multiple antenna access points, multiple candidate detector grids, at least one uplink detector terminal, and multiple served users. The access points store and update the angular power spectrum state that changes with location within the service area.

[0022] Figure 1 illustrates a complete processing chain in an indoor Wi-Fi scenario. Multiple antenna access points serve active users, and gray markers represent all candidate observation grids. During the period... Only sets The grid in the sensor array is occupied by probe terminals or portable probes. These probes transmit known uplink pilots, from which access points extract local directional power. The resulting observations are used to update the stored electromagnetic twin state; querying the updated state at active user locations yields the covariance required for downlink beamforming. The remaining uncertainty after the update is used to determine the next batch of probe grids. Therefore, probes are not an independent sensing task, but rather part of the next round and rate decision-making process.

[0023] Divide the service area into Candidate detector grids divide the angle domain into one direction .element Represents a grid In direction The average received power. Combining all elements yields the persistent electromagnetic twin state, i.e., the electromagnetic twin state matrix:

[0024] Among them, the OK It is a grid The angular power spectrum. Compared to scalar power maps, it preserves the directional information required by the antenna array; compared to instantaneous complex CSI, it is suitable for saving across update cycles.

[0025] make Indicates period The grid set selected for uplink probing is: Select the corresponding row from the angular power spectrum. Each probe contributes one row. A directional power vector. After pilot processing at the access point, the observation satisfies:

[0026] in, , To observe noise. Because electromagnetic twins preserve... Currently, sparse detection only requires estimating the amount of innovation:

[0027] Subtracting the saved predictions from the new probe data yields:

[0028] In unchanged regions, the expected innovation is zero; therefore, previously learned knowledge is preserved without having to reconstruct the entire state from a small number of current samples. (Graph Laplace matrix) This indicates the correlation between adjacent grids, allowing a local change detected at a certain location to affect nearby unobserved grids.

[0029] Step 2: Query from angular power spectrum to user covariance.

[0030] have The antenna access point is located in the grid. of Provides services to a single-antenna user. Assume the normalized array response is... The direction dictionary is: In users Location lookup for electromagnetic twins yields:

[0031] The first term superimposes the covariance components in each direction. Provides diagonal loading. The current channel satisfies... Therefore, electromagnetic twins convert sparse physical observations into beamforming knowledge through the following chain: .

[0032] Assume the data symbol power is 1, and the beam matrix is... ,user The received signal is:

[0033] in, This is received noise. Using the covariance obtained from the query as second-order channel knowledge, a moment-based rate can be constructed:

[0034] The system's operational objective is the sum of its values ​​and its velocity, not the map's normalized mean square error. For the state obtained from the current sparse probe, the access point is solved as follows:

[0035] All are determined by the query formula. .question The preserved electromagnetic twin state is given a clear wireless communication purpose: different detection observations will produce different covariance estimates, beamwidths, reachability, and rates. The remaining question is which probe should be detected next? One grid. Let This indicates a new batch of probes, This represents a new state based on the saved electromagnetic twin and future observations. Assuming the queried user remains active in the next update interval, the ideal finite budget problem is:

[0036] question It is a two-stage stochastic programming problem. The first stage is... Decision before the unknown After these measurements update the electromagnetic twin, the second-phase remedial decision is made. Maximize the results and the rate.

[0037] Step 3: Update the electromagnetic twin (electromagnetic twin state matrix) using sparse detection.

[0038] Since the angular power spectrum of the previous period has been preserved, the current observations estimate the innovation amount through the following convex quadratic problem:

[0039] in: The consistency between the control estimation results and the selected detection data; The observed changes propagate along adjacent grid lines; Encourage the status quo to be maintained; To ensure the matrix is ​​strictly positive definite, the normal equation is given:

[0040] The update also quantifies uncertainty to support future exploration. and They are respectively and The Columns, and define With noise accuracy of And a priori In this case, the innovation estimate in the formula is simultaneously the posterior mean and the maximum a posteriori probability estimate.

[0041] The posterior covariance shared by all angular components is The truncation operation ensures that the power is non-negative, and a single matrix decomposition can update all the data. Each angular component. This stage simultaneously outputs the angular power spectrum used for beamforming and the uncertainty used for selecting the probe point. In contrast, static CKM does not retain historical states, relying only on the current state. estimate .

[0042] Step 4: Maximize the sum rate based on electromagnetic twin queries.

[0043] Sparse detection determination After querying the covariance, the problem... It remains non-convex, but can be updated using a projective, successive convex approximation. Definition:

[0044] By expressing each user rate as the difference between the "logarithm of total received power" and the "logarithm of interference plus noise power," we can obtain information about the beam. Wirtinger gradient:

[0045] Combine all user gradients into In the first In the next iteration, a near-end lower bound function is used:

[0046] Among them, backtracking search gradually increases. until the candidate solution satisfies .

[0047] Maximizing this concave quadratic function within the emitter power sphere is a convex subproblem. After formulation, we obtain:

[0048] Only when the beam power exceeds The projection will only scale at that time. The compactness of the lower bound function and the backtracking condition guarantee that the sum and rate are monotonically non-decreasing; under standard MM / SCA conditions, the limit point is a stationary point. This embodiment uses the generalized eigenvector beam initialization algorithm. The projective successive convex approximation can directly present the role of electromagnetic twin covariance in optimization.

[0049] Step 5: Select a detection location that is relevant to the rate.

[0050] For discrete outer layer decision For a probe point to be valuable, it must simultaneously meet two conditions: the location has significant uncertainty and it is relevant to active users. Define candidate grids. Query relevance weights:

[0051] in, and Representing candidate grids respectively and users Location, Controlling the spatial scale of the user's relevant area. Control the query weight strength. Since the query formula is linear with respect to the angular power spectrum, and... On a compact power feasible set, the sum and rate locally satisfy Lipschitz continuity. Therefore, for some finite constant... ,have:

[0052] As can be seen, the query-weighted uncertainty provides an upper bound on the communication target error, and the following proxy problem can be constructed:

[0053] in, Is it added to batch? The innovation covariance after that. For candidate grid Using the matrix inversion lemma, we can obtain the covariance after adding a first probe:

[0054] and problems The precise single-point variance reduction:

[0055] Each round of selection Then let ,repeat Next, when the covariance matrix is ​​diagonal, the rule degenerates into query-weighted maximum variance selection; when the covariance contains off-diagonal elements, the reduction in uncertainty of the relevant grid at that probe point is also considered. Thus, the above rate sensitivity bound links each probe location to the next round and rate decision.

[0056] Therefore, this embodiment provides a sparse detection and rate-maximizing beamforming method. This method constructs an electromagnetic twin system model including multiple antenna access points, candidate detector grids, and served users, preserving the angular power spectrum state as it changes with location. Based on sparse uplink observations, graph regularization is used to estimate the innovation of the updated angular power spectrum and its posterior uncertainty. The updated angular power spectrum is queried at the location of active users to obtain the user channel covariance. Under transmit power constraints, a projective successive convex approximation is used to solve the rate-maximizing problem. Finally, the next batch of detector grids is selected based on the query-weighted posterior variance reduction criterion. This method can reuse persistent wireless knowledge under a sparse detection budget, improve multi-user rate, and couple detector selection with communication utility in a closed loop, enhancing the scalability and practicality of electromagnetic twin beamforming.

[0057] For example, a practical simulation experiment was conducted on the beamforming method provided in the above embodiments, as follows: This embodiment is in Simulation verification was performed in a scenario with a region and a total of 300 candidate grids. Access point configuration. Uniform linear array of array elements, serving One queried user; angular power spectrum used From each angle, covering The graph estimation parameters are set to... , , Diagonal loading coefficient Transmit signal-to-noise ratio 50 beam iterations were performed at dB. An average of 20 noise tests were conducted at each experimental point, with 25 channel samples used in each noise test, for a total of 500 common channels.

[0058] Baseline methods include: perfect covariance, electromagnetic twin, static CKM, and stale twin. All methods employ the same beam solver to isolate the value of propagating knowledge itself.

[0059] like Figure 2 As shown, when only 1% of the grid is randomly probed, electromagnetic twin beamforming reaches 2.688 bit / s / Hz, while static CKM only reaches 1.116 bit / s / Hz, indicating that persistent state can significantly improve the rate at extremely low probe density.

[0060] like Figure 3 As shown, when the detection ratio is 1%, the normalized mean square error of the angular power spectrum of the electromagnetic twin is -7.51 dB, and the static CKM is -0.40 dB; when the detection ratio is 12%, the electromagnetic twin improves to -10.02 dB. This indicates that the electromagnetic twin can combine sparse new evidence with previously learned knowledge.

[0061] like Figure 4 As shown, when the query sensing detection position is fixed at 7% and the transmit signal-to-noise ratio is changed, the electromagnetic twin reaches 6.521 bit / s / Hz under 20 dB conditions, which is 26.1% higher than the old twin's 5.172 bit / s / Hz. This indicates that the current direction propagation knowledge can significantly reduce multi-user leakage interference.

[0062] like Figure 5 As shown, with a cumulative detection of 7% of the grid, rate-dependent selection enables electromagnetic twinning and rate to reach 3.033 bit / s / Hz, which is only 3.4% different from perfect covariance beamforming; random detection reaches 2.695 bit / s / Hz, which is 12.6% lower with the same observation budget.

[0063] like Figure 6 As shown, the normalized mean square error of the map for rate-correlated detection is -10.24 dB, which is better than -9.36 dB for random detection, and also better than old twins and static CKM. This demonstrates that sparse detection reduces the problem... The updated electromagnetic twin improves user covariance by reducing uncertainty, and the more accurate covariance further improves and maximizes the rate of results.

[0064] For example, such as Figure 8 As shown, this embodiment also provides a sparse detection and sum rate maximizing beamforming system, including: The model building module is used to construct an electromagnetic twin system model, which includes a set of candidate detector grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detector observation model, and a graph Laplace matrix; the graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detector grids. The state update module is used to obtain the uplink detection observation data corresponding to the selected detection grid set in the current period based on the electromagnetic twin system model, estimate the state innovation amount in the current period by combining the electromagnetic twin state matrix saved in the previous period, update the electromagnetic twin state matrix in the current period, and calculate the posterior covariance matrix corresponding to the current electromagnetic twin state. The covariance query module is used to query the updated current period electromagnetic twin state matrix based on the grid location of each active user, and generate the channel covariance matrix corresponding to each active user. The beam optimization module is used to solve the downlink sum and rate maximization problem under the constraint of total transmit power, based on the channel covariance matrix of each active user, to obtain the optimal downlink beamforming matrix. The detector selection module combines the posterior covariance matrix and the query-related weights of each candidate detector grid relative to active users. With the goal of minimizing the upper bound of the sum rate estimation error, it selects the next round of candidate detector grids from the undetected candidate detector grids and feeds the next round of candidate detector grids back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

[0065] The present invention also provides a sparse detection and sum rate maximizing beamforming device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the sparse detection and sum rate maximizing beamforming method.

[0066] The present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the sparse detection and sum rate maximizing beamforming method.

[0067] When the processor executes the computer program, it implements the above-mentioned steps of sparse detection and rate-maximizing beamforming, for example: constructing an electromagnetic twin system model, which includes a set of candidate detection grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detection observation model, and a graph Laplace matrix; the graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detection grids. Based on the electromagnetic twin system model, the uplink detection observation data corresponding to the selected detection grid set in the current period is obtained. The state innovation quantity of the current period is estimated by combining the electromagnetic twin state matrix saved in the previous period, and the electromagnetic twin state matrix of the current period is updated. At the same time, the posterior covariance matrix corresponding to the current electromagnetic twin state is calculated. Based on the grid location of each active user, query the updated current period electromagnetic twin state matrix to generate the channel covariance matrix for each active user. Based on the channel covariance matrix of each active user, the downlink sum rate maximization problem is solved under the constraint of total transmit power to obtain the optimal downlink beamforming matrix. By combining the posterior covariance matrix and the query-related weights of each candidate probe grid relative to active users, and with the goal of minimizing the upper bound of the sum rate estimation error, a set of candidate probe grids for the next round is selected from the undetected candidate probe grids. The set of candidate probe grids for the next round is fed back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

[0068] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing preset functions, the instruction segments describing the execution process of the computer program in the sparse detection and rate-maximizing beamforming device. For example, the computer program can be divided into a model building module, a state update module, a covariance query module, a beam optimization module, and a detector selection module; the specific functions are as follows: The model building module is used to build an electromagnetic twin system model, which includes a set of candidate detector grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detection observation model, and a Graph Laplace matrix; the Graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detector grids; The state update module is used to obtain the uplink detection observation data corresponding to the selected detector grid set in the current period based on the electromagnetic twin system model, estimate the state innovation amount in the current period by combining the electromagnetic twin state matrix saved in the previous period, update the electromagnetic twin state matrix in the current period, and simultaneously calculate the current period's state innovation amount. The system comprises the following modules: a posterior covariance matrix corresponding to the previous electromagnetic twin state; a covariance query module, used to query the updated current-cycle electromagnetic twin state matrix based on the grid location of each active user, generating the channel covariance matrix for each active user; a beam optimization module, used to solve the downlink sum and rate maximization problem under the constraint of total transmit power based on the channel covariance matrix of each active user, to obtain the optimal downlink beamforming matrix; and a detector selection module, used to select the next round of candidate detector grids from the undetected candidate detector grids, with the goal of minimizing the upper bound of the sum and rate estimation error, by combining the posterior covariance matrix and the query-related weights of each candidate detector grid relative to the active user, and feeding the next round of candidate detector grids back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

[0069] The sparse detection and sum rate maximizing beamforming device can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The sparse detection and sum rate maximizing beamforming device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above are examples of sparse detection and sum rate maximizing beamforming devices and do not constitute a limitation on such devices. It may include more components than described above, or combine certain components, or different components. For example, the sparse detection and sum rate maximizing beamforming device may also include input / output devices, network access devices, buses, etc.

[0070] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or any conventional processor. The processor is the control center of the sparse detection and sum rate maximization beamforming device, connecting various parts of the device via various interfaces and lines.

[0071] The memory can be used to store the computer program and / or modules. The processor implements various functions of the sparse detection and rate-maximizing beamforming device by running or executing the computer program and / or modules stored in the memory and calling the data stored in the memory.

[0072] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function (such as sound playback, image playback, etc.). The data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital cards (SD cards), flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0073] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the sparse detection and sum rate maximizing beamforming method described above.

[0074] If the modules / units of the sparse detection and rate-maximizing beamforming system are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0075] Based on this understanding, the present invention can implement all or part of the processes in the above-described sparse detection and sum rate maximizing beamforming method, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-described sparse detection and sum rate maximizing beamforming method. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or a preset intermediate form, etc.

[0076] The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0077] It should be noted that the content contained in the computer-readable storage medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0078] Compared with existing beamforming methods, this invention provides a sparse detection and sum rate maximization beamforming method and related apparatus, which has the following advantages: First, this invention uses the angular power spectrum as the persistent electromagnetic twin state, which not only preserves the directional information required by the array, but also avoids the problem that the instantaneous complex channel map changes too quickly and is difficult to save across cycles. Thus, it can update local innovations with a small amount of current detection and reuse historical knowledge.

[0079] Secondly, this invention obtains the user channel covariance through angular power spectrum lookup and directly uses the covariance for rate-maximizing beamforming, so that the map state has a clear communication purpose, rather than just pursuing the minimization of map reconstruction error.

[0080] Third, this invention uses the projection successive convex approximation to directly utilize the covariance moment rate target for beam optimization, ensuring that the sum rate remains monotonically constant under power constraints, and clearly demonstrating the role of electromagnetic twin covariance in optimization.

[0081] Fourth, this invention derives the rate sensitivity bound, transforming the selection of the next batch of probes into a query-weighted posterior variance reduction problem, thereby concentrating the limited probe budget on the positions most valuable to active users and rates, and achieving closed-loop coupling between sparse probes and communication utility.

[0082] The above embodiments are merely one of the implementation methods for achieving the technical solution of the present invention. The scope of protection claimed by the present invention is not limited to this embodiment, but also includes any variations, substitutions and other implementation methods that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A sparse detection and sum rate maximization beamforming method, characterized in that, include: An electromagnetic twin system model is constructed, which includes a set of candidate detector grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detector observation model, and a graph Laplace matrix; the graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detector grids. Based on the electromagnetic twin system model, the uplink detection observation data corresponding to the selected detection grid set in the current period is obtained. The state innovation quantity of the current period is estimated by combining the electromagnetic twin state matrix saved in the previous period, and the electromagnetic twin state matrix of the current period is updated. At the same time, the posterior covariance matrix corresponding to the current electromagnetic twin state is calculated. Based on the grid location of each active user, query the updated current period electromagnetic twin state matrix to generate the channel covariance matrix for each active user. Based on the channel covariance matrix of each active user, the downlink sum rate maximization problem is solved under the constraint of total transmit power to obtain the optimal downlink beamforming matrix. By combining the posterior covariance matrix and the query-related weights of each candidate probe grid relative to active users, and with the goal of minimizing the upper bound of the sum rate estimation error, a set of candidate probe grids for the next round is selected from the undetected candidate probe grids. The set of candidate probe grids for the next round is fed back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

2. The sparse detection and sum rate maximization beamforming method according to claim 1, characterized in that, The construction of the electromagnetic twin system model includes: The service area is divided into N candidate detection grids, forming a candidate detection grid set; The angle domain is divided into L directions, forming a set of angle domain directions; Construct an electromagnetic twin state matrix of dimension N×L; wherein, the angular power spectrum of the candidate detector grid is: the element in the nth row of the electromagnetic twin state matrix corresponds to the average received power of the nth candidate detector grid in each direction; Define a probe selection matrix; the probe selection matrix is ​​used to select the row vector corresponding to the currently probed grid from the electromagnetic twin state matrix; An uplink sparse detection observation model is established based on the detector selection matrix; the observed value of the uplink sparse detection observation model is equal to the product of the detector selection matrix and the electromagnetic twin state matrix plus the observation noise. A graph Laplacian matrix is ​​constructed to characterize the spatial power correlation between adjacent candidate detector grids, so that locally detected power changes can propagate to adjacent undetected candidate detector grids.

3. The sparse detection and sum rate maximization beamforming method according to claim 1, characterized in that, The process involves estimating the state innovation of the current period by combining the electromagnetic twin state matrix stored from the previous period, updating the electromagnetic twin state matrix of the current period, and simultaneously calculating the posterior covariance matrix corresponding to the current electromagnetic twin state, including: The observation residual matrix is ​​obtained by subtracting the product of the current cycle's uplink detection observation data and the detection selection matrix and the electromagnetic twin state matrix saved in the previous cycle. A convex optimization objective function is constructed, which includes a data fitting term, a spatial smoothing term, and a state preservation term. The data fitting term is used to constrain the consistency between the state innovation quantity and the observation residual, the spatial smoothing term realizes the spatial propagation constraint of power change based on the graph Laplacian matrix, and the state preservation term is used to constrain the overall magnitude of the state innovation quantity. The convex optimization objective function is solved based on the normal equation to obtain the state innovation matrix of the current period. The electromagnetic twin state matrix of the previous period is added to the state innovation matrix and then truncated non-negatively to obtain the electromagnetic twin state matrix of the current period. At the same time, the posterior covariance matrix shared by all angular components is solved. The posterior covariance matrix is ​​used to quantify the estimation uncertainty of the current electromagnetic twin state.

4. The sparse detection and sum rate maximizing beamforming method according to claim 1, characterized in that, The generation of the channel covariance matrix for each active user includes: Construct a direction dictionary matrix consisting of the normalized array response vectors corresponding to each direction; For each active user, the angular power spectrum row vector corresponding to the candidate detection grid where the active user is located is extracted from the electromagnetic twin state matrix of the current period, and the initial channel covariance matrix is ​​calculated by combining it with the direction dictionary matrix. A diagonal loading process is applied to the initial channel covariance matrix, with the diagonal loading amount being proportional to the trace of the initial channel covariance matrix, to obtain the final user channel covariance matrix; wherein, the user channel covariance matrix is ​​used to characterize the second-order statistical characteristics of the channel at the corresponding candidate probe grid location.

5. The sparse detection and sum rate maximization beamforming method according to claim 1, characterized in that, The problem of maximizing downlink sum and rate under the constraint of total transmit power includes: Based on the channel covariance matrix of each active user, the reachable rate expression for each active user is constructed, and the reachable rates of all active users are superimposed to obtain the system and rate objective functions. Calculate the Wirtinger gradients of the system and rate objective functions with respect to the beamforming matrix, and construct the near-end lower bound concave quadratic functions of the system and rate. The iteration step size parameter is determined by backtracking search. The optimal solution is obtained for the concave quadratic function of the near-end lower bound under the constraint of total transmit power. The solution is then projected onto the feasible region of transmit power to obtain the iteratively updated beamforming matrix. Repeat the iterations until the system and rate converge, and obtain the final optimal downlink beamforming matrix.

6. The sparse detection and sum rate maximizing beamforming method according to claim 1, characterized in that, The selection of the next round of candidate detector grids from the undetected candidate detector grids includes: Based on the spatial distance between each candidate probe grid and each active user, the query-related weights of each candidate probe grid are calculated using the Gaussian kernel function, and a query weight diagonal matrix is ​​constructed. With minimizing the query weighted posterior uncertainty as the proxy optimization objective, the candidate probe grid with the largest single-point variance reduction is selected sequentially from the undetected candidate probe grids until the number of selected candidate probe grids reaches the preset single-round detection limit, thus obtaining the candidate probe grid set for the next round. The agent optimization objective corresponds to the upper bound of the system and rate estimation error.

7. The sparse detection and sum rate maximizing beamforming method according to claim 6, characterized in that, The process of sequentially selecting the candidate probe grid that maximizes the reduction in query weighted uncertainty from the unexplored candidate probe grids includes: Using the matrix inversion lemma, we can calculate the updated posterior covariance matrix corresponding to the addition of a single candidate detector grid to the detector set. Based on the query weight diagonal matrix and the posterior covariance matrix before and after the update, the single-point variance reduction corresponding to the candidate probe grid is calculated.

8. A sparse detection and sum rate maximizing beamforming system, characterized in that, include: The model building module is used to construct an electromagnetic twin system model, which includes a set of candidate detector grids after partitioning, a set of angular domain directions, an electromagnetic twin state matrix based on angular power spectrum, an uplink sparse detector observation model, and a graph Laplace matrix; the graph Laplace matrix is ​​used to characterize the spatial power correlation between adjacent candidate detector grids. The state update module is used to obtain the uplink detection observation data corresponding to the selected detection grid set in the current period based on the electromagnetic twin system model, estimate the state innovation amount in the current period by combining the electromagnetic twin state matrix saved in the previous period, update the electromagnetic twin state matrix in the current period, and calculate the posterior covariance matrix corresponding to the current electromagnetic twin state. The covariance query module is used to query the updated current period electromagnetic twin state matrix based on the grid location of each active user, and generate the channel covariance matrix corresponding to each active user. The beam optimization module is used to solve the downlink sum and rate maximization problem under the constraint of total transmit power, based on the channel covariance matrix of each active user, to obtain the optimal downlink beamforming matrix. The detector selection module combines the posterior covariance matrix and the query-related weights of each candidate detector grid relative to active users. With the goal of minimizing the upper bound of the sum rate estimation error, it selects the next round of candidate detector grids from the undetected candidate detector grids and feeds the next round of candidate detector grids back to the electromagnetic twin system model for uplink sparse detection and state update in the next cycle.

9. A sparse detection and sum rate maximizing beamforming device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the sparse detection and rate-maximizing beamforming method according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it is used to implement the sparse detection and rate-maximizing beamforming method according to any one of claims 1-7.

Images