Compression matrix design method for uplink non-cellular access system

Abstract

The invention provides a compression matrix design method for an uplink non-cellular access system, which belongs to the technical field of wireless communication, and comprises the following steps: generating a plurality of non-cellular MIMO (Multiple Input Multiple Output) system scenes in advance; the method comprises the following steps: performing task sampling from a pre-generated cellular-free MIMO system scene to obtain a plurality of subtasks, constructing a joint task by the plurality of subtasks, and modeling the cellular-free MIMO system scene corresponding to each subtask as a heterogeneous graph model; performing joint training on the joint task based on the meta-learning parameters, generating a compression matrix corresponding to each sub-task, obtaining an optimal parameter corresponding to the joint task, and adopting a message passing algorithm in the joint training process; calculating a difference vector between the optimal parameter and the meta-learning parameter to complete updating of the meta-learning parameter, and performing next joint training to finally obtain the optimal meta-learning parameter; and deploying the optimal meta-learning parameter in a cellular-free MIMO system following a standard block fading model, and performing data transmission. The method is suitable for a dynamic communication scene.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a compression matrix design method for an uplink non-cellular access system. Background Art

[0002] In recent years, the booming development of mobile internet and the Internet of Things (IoT) has placed higher demands on wireless communication systems, including data transmission rates, network coverage, and system capacity. While traditional cellular network architecture has evolved over the years, it still faces significant challenges in coping with the massive access of devices and high-traffic traffic. To overcome the inherent limitations of this architecture, cell-free massive MIMO (Multiple-Input Multiple-Output) technology has garnered widespread attention as an innovative solution. This technology utilizes a large number of distributed access points (APs) connected to a central processing unit (CPU) via fronthaul links for coordinated signal processing, thereby providing more uniform network coverage and higher spectral efficiency. This architectural innovation effectively addresses the coverage holes and capacity bottlenecks in traditional cellular networks, providing a new technical path to achieving the performance goals of next-generation communication systems. However, in cell-free MIMO systems, the fronthaul links between the APs and the CPU are typically capacity-constrained. This limitation hinders the system's ability to effectively and timely process the large number of parallel uplink signals when faced with the concurrent access of a massive number of devices, resulting in reduced spectral efficiency. Therefore, in order to address the capacity limitation of the fronthaul link between the access point and the central processing unit, each access point needs to compress the uplink signal in advance when transmitting it to the central processing unit to reduce the fronthaul overhead.

[0003] To ensure information integrity during signal compression, existing research typically employs signal dimensionality reduction methods based on compression matrices. However, in non-cellular MIMO systems, the distributed architecture of multiple access points working together makes the joint optimization of the compression matrix particularly complex. The widespread application of deep learning offers an effective solution to this problem. Existing compression matrix design methods based on deep neural networks (DNNs) can perceive the effective information contained in the uplink signal from each access point and design the compression matrix for each access point with the goal of improving system capacity. This compression matrix stores the effective information in the uplink signal in a low-dimensional form, thereby reducing fronthaul communication overhead. However, existing DNN-based compression matrix design methods typically rely on fixed model parameters for training. In actual deployment, when system parameters change, such as the number of users, the original model parameters become inappropriate. This scalability limitation makes these DNN-based compression matrix design methods difficult to apply to dynamic communication systems. Summary of the Invention

[0004] The purpose of the present invention is to propose a compression matrix design method for an uplink non-cellular access system, which is suitable for dynamic communication scenarios.

[0005] The present invention is achieved through the following technical solutions: A compression matrix design method for an uplink non-cellular access system includes the following steps: Step S1: Pre-generate multiple non-cellular MIMO system scenarios, each of which includes the same number of access points, a different number of users, different user location distributions, and different rate requirements; Step S2: Sample tasks from a pre-generated non-cellular MIMO system scenario to obtain several subtasks, and construct a joint task from these subtasks. Each subtask contains a different number of users, user distribution, and rate requirements. Based on a graph neural network, the non-cellular MIMO system scenario corresponding to each subtask is modeled as a heterogeneous graph model. Step S3: Jointly train the joint task based on the learning parameters to generate a compression matrix corresponding to each subtask and obtain the optimal parameters corresponding to the joint task. The joint training process uses a message passing algorithm, and the meta-learning parameters for the first joint training are randomly generated. Step S4: Calculate the difference vector between the optimal parameters and the meta-learning parameters, and complete the update of the meta-learning parameters with a fixed step size and the difference vector, and then enter step S1 for the next joint training to continuously optimize the meta-learning parameters; Step S5: deploy the optimized meta-learning parameters in a cell-free MIMO system that follows a standard block fading model and perform data transmission.

[0006] Furthermore, in step S2, the nodes in the heterogeneous graph model are users and access points, the edges are the connection relationships between access points and users, the characteristics of users are rate requirements, the characteristics of access points are the compression matrix to be optimized, and the characteristics of edges are the uplink channel matrix between the connected access points and users.

[0007] Furthermore, the message passing algorithm in step S3 includes feature definition and message passing, specifically: Feature definition: For each subtask in the joint task, the compression matrix of each access point corresponding to it is initialized with a random number to obtain the initialized compression matrix , and based on the channel parameters between the user and the access point and user rate requirements Complete the feature definition of nodes and edges in the corresponding heterogeneous graph model, where: l Indicates the access point number. Indicates the subtasks, k Indicates the k users; Message delivery includes message aggregation and feature updates, specifically: Message aggregation: the first l An access point obtains information from all its neighboring nodes. The process is expressed as ,in, Indicates the The corresponding subtask l The access point is t The information obtained during round optimization, is the pooling function, For the l The set of all neighbor nodes of an access point, For the parameter Parameterized functions of drivers; Feature Update: l Each access point completes the update of the features in the next round of optimization based on the acquired messages and the features in the current optimization round. The process is expressed as ,in, For the The corresponding subtask l Access point t Features during round optimization, For the parameter Parameterized function of the driver.

[0008] Furthermore, in step S3, in each round of optimization, each access point completes message transmission in turn and performs T Round optimization to obtain The final compression matrix under subtasks , based on the known channel state information between each access point and the user and the final compression matrix Complete the collaborative processing of the uplink signal. k The signal-to-interference-noise ratio of a user during transmission is .

[0009] Furthermore, in step S3, after obtaining the final compression matrix of all subtasks, according to the formula Calculate the performance loss of the current graph neural network on the joint task , repeat the joint training until Converge and obtain the optimal parameters for the joint task , where Ω represents the number of subtasks, For the The number of users for each subtask, For the User rate requirements for each subtask, λ is the penalty coefficient.

[0010] Furthermore, in step S4, according to the formula Complete the update of meta-learning parameters, is a fixed step size.

[0011] Furthermore, the step S5 specifically includes the following steps: Step S51: At the beginning of each channel block, each access point estimates the user's uplink channel parameters and transmits them to the central processing unit via the fronthaul link. The central processing unit completes the construction of a heterogeneous graph model corresponding to the non-cellular MIMO system scenario of the current channel block based on the information uploaded by all access points. Step S52: The central processing unit uses the meta-learning parameters after offline optimization to initialize the parameters of the graph neural network in the heterogeneous graph model. Based on the current heterogeneous graph model, the central processing unit completes the feature definition and message passing process of each access point in parallel to optimize the compression matrix corresponding to each access point. In step S53, the central processing unit transmits the optimized compression matrix back to the corresponding access point. During the remaining time of the channel block, each access point receives the user uplink data, compresses the user uplink data based on the optimized compression matrix, and transmits the compressed data to the central processing unit. The central processing unit uses the minimum mean square error algorithm based on the optimized compression matrix and the system channel state information to complete the coordinated processing of each user's uplink data.

[0012] Furthermore, in each training session, several subtasks are sampled from different non-cellular MIMO system scenarios to construct a joint task.

[0013] The present invention has the following beneficial effects: 1. The present invention pre-generates multiple non-cellular MIMO system scenarios. Before each joint training, tasks are sampled from the pre-generated non-cellular MIMO system scenarios to obtain several subtasks. A joint task is constructed from these subtasks. The joint task is then jointly trained based on learning parameters to generate a compression matrix corresponding to each subtask and obtain the optimal parameters for the joint task. The meta-learning parameters are then updated by calculating the difference vector between the optimal parameters and the meta-learning parameters using a fixed step size and the difference vector. The joint training is repeated multiple times to continuously optimize the meta-learning parameters. Finally, the optimized meta-learning parameters are deployed in a non-cellular MIMO system that follows a standard block counting model and data is transmitted. The compression matrix designed in this way can adapt to dynamic changes in the system. Even if the number of users in the scenario changes, it can still quickly adapt to the new communication environment, meeting the high-quality, high-speed service requirements of the dynamic system. In this process, the non-cellular MIMO system is modeled as a heterogeneous graph structure and combined with a message passing algorithm to effectively solve the problem of model failure caused by dynamic changes in the number of users. The introduction of meta-learning technology improves the adaptability and generalization ability of the model in dynamic scenarios by dividing the task scenarios and conducting joint training. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described in detail below with reference to the accompanying drawings.

[0015] Figure 1 Flowchart of the present invention.

[0016] Figure 2 The non-cellular MIMO system scenario of the present invention is modeled as a schematic diagram of a heterogeneous graph model.

[0017] Figure 3 It is a flow chart of the online optimization and actual transmission stages of the present invention.

[0018] Figure 4 The figure is a performance comparison chart of the present invention and the comparison algorithm at different compression dimensions. DETAILED DESCRIPTION

[0019] like Figure 1 As shown, the compression matrix design method for uplink non-cellular access system includes the following steps: Step S1: Pre-generate multiple non-cellular MIMO system scenarios, each of which includes the same number of access points, a different number of users, different user location distributions, and different rate requirements, to simulate a real dynamic communication environment; The number of users in the same type of non-cellular MIMO system scenario is the same, but the user's location distribution and rate requirements are different. Specifically in this embodiment, the number of access points in each non-cellular MIMO system scenario is L Both are 4, number of antennas M Both are 12, the equivalent number of antennas after compression N Both are 6, number of users K ∈{8,12,16,20}, each user is randomly distributed within the service range.

[0020] Step S2: Sample tasks from a pre-generated non-cellular MIMO system scenario to obtain several subtasks, and construct a joint task from these subtasks. Each subtask contains a different number of users, user distribution, and rate requirements. Based on a graph neural network, the non-cellular MIMO system scenario corresponding to each subtask is modeled as a heterogeneous graph model. like Figure 2 As shown, the construction of a heterogeneous graph model involves node construction and edge construction. Node construction involves constructing different access points and users into two different types of nodes in the heterogeneous graph model: access points and users. User characteristics include their respective rate requirements, and access point characteristics are the compression matrix to be optimized. Edge construction involves constructing the connections between access points and users into edges in the heterogeneous graph model. Edge characteristics are the uplink channel matrix between the connected access points and users. Each subtask in the joint task is modeled as a unique heterogeneous graph model based on its specific scenario distribution.

[0021] In heterogeneous graph models, user and edge features serve as input for compression matrix optimization and remain constant during the optimization process. Access point features contain the compression matrix parameters to be optimized. To efficiently process system input information, the compression matrix optimization process employs a message passing algorithm.

[0022] Step S3: Jointly train the joint task based on the meta-learning parameters, generate the compression matrix corresponding to each sub-task, and obtain the optimal parameters corresponding to the joint task. The joint training process adopts the message passing algorithm. A set of meta-learning parameters for the first joint training is Random generation; The message passing algorithm includes feature definition and message passing, specifically: Feature definition: For each subtask in the joint task, the compression matrix of each access point corresponding to it is initialized with a random number to obtain the initialized compression matrix , and based on the channel parameters between the user and the access point and user rate requirements Complete the feature definition of nodes and edges in the corresponding heterogeneous graph model, where: l Indicates the access point number. Indicates the subtasks, k Indicates the k users; Message delivery includes sequential message aggregation and feature updates, specifically: Message aggregation: the first l An access point obtains information from all its neighboring nodes. The process is expressed as ,in, Indicates the The corresponding subtask l The access point is t The information obtained during round optimization, is the pooling function, For the l The set of all neighbor nodes of an access point, For the parameter Parameterized functions of drivers; Feature Update: l Each access point completes the update of the features in the next round of optimization based on the acquired messages and the features in the current optimization round. The process is expressed as ,in, For the The corresponding subtask l Access point t Features during round optimization, For the parameter Driven parameterized function, meta-learning parameters include parameters and parameters ; In each round of optimization, each access point completes message aggregation and feature update in turn, and T Round optimization (i.e. repeated message aggregation and feature update process T times), to obtain the The final compression matrix under subtasks The central processing unit is based on the minimum mean square error algorithm, according to the known channel state information between each access point and the user and the final compression matrix Complete the collaborative processing of the uplink signal. k The signal-to-interference-noise ratio of a user during transmission is , wherein the process of obtaining the channel state information is the existing technology; After obtaining the final compression matrix of all subtasks, according to the formula Computing the performance loss of graph neural networks on joint tasks , repeat the joint training until Converge and obtain the optimal parameters of the graph neural network for this joint task , where Ω represents the number of subtasks, For the The number of users for each subtask, For the User rate requirements for each subtask, λ is the penalty coefficient.

[0023] To ensure the wide adaptability of the training process, during each training session, several subtasks are sampled from different non-cellular MIMO system scenarios to construct a joint task. Each subtask contains a different number of users, location distribution, and rate requirements, allowing the model to learn diverse scenario characteristics.

[0024] Step S4: Calculate the difference vector between the optimal parameters and the meta-learning parameters, and complete the update of the meta-learning parameters with a fixed step size and the difference vector, and then enter step S1 for the next joint training to continuously optimize the meta-learning parameters; Specifically, according to the formula Complete the update of meta-learning parameters, is a fixed step size.

[0025] After multiple rounds of task sampling, joint training, and parameter optimization, the meta-learning parameter θ will contain prior information from different scenarios during training. When faced with new scenarios, the model can quickly adapt to unknown system parameters, thereby reducing the latency during the compression matrix optimization process and improving the final performance of the optimized compression matrix.

[0026] Step S5: deploying the optimized meta-learning parameters in a non-cellular MIMO system following a standard block fading model and performing data transmission; The specific steps include: Step S51: At the beginning of each channel block, each access point estimates the user's uplink channel parameters and transmits them to the central processing unit via the fronthaul link. The central processing unit completes the construction of a heterogeneous graph model corresponding to the non-cellular MIMO system scenario of the current channel block based on the information uploaded by all access points. In this embodiment, a cell-free MIMO system that follows the standard block fading model includes four access points with 12 antennas and one central processing unit. The access points are deployed in a square grid pattern. The number of single-antenna users changes dynamically between different channel blocks, and the users are randomly distributed. To reduce the communication overhead of the fronthaul link, it is necessary to reduce the antenna dimension of the access point to 6 by compressing the matrix.

[0027] Step S52: The central processing unit uses the meta-learning parameters after offline optimization to initialize the parameters of the graph neural network in the heterogeneous graph model, and completes the feature definition and message passing process of each access point in parallel based on the current heterogeneous graph model to complete the optimization of the compression matrix corresponding to each access point. During the optimization process, the central processing unit performs multiple rounds of updates to the model parameters based on the current scenario. Because the meta-learning parameters fully extract key features of compression matrix design in non-cellular MIMO systems during training and incorporate prior information from multiple different scenarios, the model can still achieve excellent performance with a limited number of updates, even under extremely strict latency constraints during the actual online optimization process.

[0028] In step S53, the central processing unit transmits the optimized compression matrix back to the corresponding access point. During the remaining time of the current channel block, each access point receives the user uplink data, compresses the user uplink data based on the optimized compression matrix, and transmits the compressed data to the central processing unit to reduce the communication overhead of the fronthaul link. The central processing unit uses the minimum mean square error algorithm based on the optimized compression matrix and system channel state information to complete the coordinated processing of each user's uplink data.

[0029] The flow chart of online optimization and actual transmission phase is as follows: Figure 3 shown.

[0030] Figure 4 The figure shows a performance comparison between the present invention and the comparison algorithm (antenna selection algorithm) at different compression dimensions. It can be seen that, compared with the antenna selection algorithm, when the compression dimension is low, the present invention exhibits good compression performance and can effectively retain the original information in the compressed low-dimensional representation. At the same time, when the compression dimension is high, the present invention can achieve performance close to that of no compression, and can effectively reduce the communication overhead between the access point and the central processing unit while ensuring near-lossless performance.

[0031] The above description is merely a preferred embodiment of the present invention and therefore cannot be used to limit the scope of the present invention. In other words, equivalent changes and modifications made according to the scope of the patent application and the contents of the specification should still fall within the scope of the patent of the present invention.

Claims

1. A compression matrix design method for an uplink non-cellular access system, characterized by: The steps include: Step S1: Pre-generate multiple non-cellular MIMO system scenarios, each of which includes the same number of access points, a different number of users, different user location distributions, and different rate requirements; Step S2: Sample tasks from a pre-generated non-cellular MIMO system scenario to obtain several subtasks, and construct a joint task from these subtasks. Each subtask contains a different number of users, user distribution, and rate requirements. Based on a graph neural network, the non-cellular MIMO system scenario corresponding to each subtask is modeled as a heterogeneous graph model. Step S3: Jointly train the joint task based on the learning parameters to generate a compression matrix corresponding to each subtask and obtain the optimal parameters corresponding to the joint task. The joint training process uses a message passing algorithm, and the meta-learning parameters for the first joint training are randomly generated. Step S4: Calculate the difference vector between the optimal parameters and the meta-learning parameters, and complete the update of the meta-learning parameters with a fixed step size and the difference vector, and then enter step S1 for the next joint training to continuously optimize the meta-learning parameters; Step S5: deploy the optimized meta-learning parameters in a cell-free MIMO system that follows a standard block fading model and perform data transmission.

2. The method for designing a compression matrix for an uplink non-cellular access system according to claim 1, wherein: In step S2, the nodes in the heterogeneous graph model are users and access points, the edges are the connection relationships between access points and users, the characteristics of users are rate requirements, the characteristics of access points are the compression matrix to be optimized, and the characteristics of edges are the uplink channel matrix between the connected access points and users.

3. The method for designing a compression matrix for an uplink non-cellular access system according to claim 2, wherein: The message passing algorithm in step S3 includes feature definition and message passing, specifically: Feature definition: For each subtask in the joint task, the compression matrix of each access point corresponding to it is initialized with a random number to obtain the initialized compression matrix , and based on the channel parameters between the user and the access point and user rate requirements Complete the feature definition of nodes and edges in the corresponding heterogeneous graph model, where: l Indicates the access point number. Indicates the subtasks, k Indicates the k users; Message delivery includes message aggregation and feature updates, specifically: Message aggregation: the first l An access point obtains information from all its neighboring nodes. The process is expressed as ,in, Indicates the The corresponding subtask l The access point is t The information obtained during round optimization, is the pooling function, For the l The set of all neighbor nodes of an access point, For the parameter Parameterized functions of drivers; Feature Update: l Each access point completes the update of the features in the next round of optimization based on the acquired messages and the features in the current optimization round. The process is expressed as ,in, For the The corresponding subtask l Access point t Features during round optimization, For the parameter Parameterized function of the driver.

4. The method for designing a compression matrix for an uplink non-cellular access system according to claim 3, wherein: In step S3, in each round of optimization, each access point completes message transmission in turn and performs T Round optimization to obtain The final compression matrix under subtasks , based on the known channel state information between each access point and the user and the final compression matrix Complete the collaborative processing of the uplink signal. k The signal-to-interference-noise ratio of a user during transmission is .

5. The method for designing a compression matrix for an uplink non-cellular access system according to claim 4, wherein: In step S3, after obtaining the final compression matrix of all subtasks, according to the formula Computing the performance loss of graph neural networks on joint tasks , repeat the joint training until Converge and obtain the optimal parameters for the joint task , where Ω represents the number of subtasks, For the The number of users for each subtask, For the The user rate requirement of each subtask, λ is the penalty coefficient.

6. The method for designing a compression matrix for an uplink non-cellular access system according to claim 5, wherein: In step S4, according to the formula Complete the update of meta-learning parameters, is a fixed step size.

7. The method for designing a compression matrix for an uplink non-cellular access system according to claim 6, wherein: The step S5 specifically includes the following steps: Step S51: At the beginning of each channel block, each access point estimates the user's uplink channel parameters and transmits them to the central processing unit via the fronthaul link. The central processing unit completes the construction of a heterogeneous graph model corresponding to the non-cellular MIMO system scenario of the current channel block based on the information uploaded by all access points. Step S52: The central processing unit uses the meta-learning parameters after offline optimization to initialize the parameters of the graph neural network in the heterogeneous graph model. Based on the current heterogeneous graph model, the central processing unit completes the feature definition and message passing process of each access point in parallel to optimize the compression matrix corresponding to each access point. In step S53, the central processing unit transmits the optimized compression matrix back to the corresponding access point. During the remaining time of the channel block, each access point receives the user uplink data, compresses the user uplink data based on the optimized compression matrix, and transmits the compressed data to the central processing unit. The central processing unit uses the minimum mean square error algorithm based on the optimized compression matrix and the system channel state information to complete the coordinated processing of each user's uplink data.

8. A compression matrix design method for an uplink non-cellular access system according to any one of claims 1 to 7, characterized in that: In each training session, several subtasks are sampled from different non-cellular MIMO system scenarios to construct a joint task.

Images