Wireless communication system and method

Abstract

The invention provides a wireless communication system and method, and the system comprises a transmitting end which is used for obtaining a wireless signal and target channel state information, and the target channel state information provides basic data for the optimal configuration of a reconfigurable intelligent surface, and the sending end is further used for optimizing the target channel state information according to the target optimization model to obtain an RIS configuration instruction more conforming to the current state, so that the reconfigurable intelligent surface adjusts a reflection unit of the reconfigurable intelligent surface in real time based on the RIS configuration instruction, and then the reconfigurable intelligent surface reflects a wireless signal, so that the wireless signal can be transmitted to the reconfigurable intelligent surface. A more reliable reflection signal which has real-time performance and conforms to the current environment state is obtained; and the user equipment is used for generating an uplink signal according to the reflected signal and sending the uplink signal to the sending end. Therefore, the requirement of a dynamically changing wireless environment is met, and the reliability of wireless communication is also improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication, and more particularly to a wireless communication system and method. Background Art

[0002] With the rapid development of information technology, mobile communication systems have evolved rapidly from the first generation to the fifth generation and are moving towards the sixth generation of wireless communication technology. 5G networks, with their high data rate, low latency, and large-scale connection capabilities, are widely used in fields such as mobile Internet, Internet of Things (IoT), intelligent manufacturing, and autonomous driving. However, with the continuous growth of user demands and the increasing complexity of application scenarios, 5G networks face many challenges, such as tight spectrum resources, high energy consumption, limited coverage, and signal interference. To address these challenges, researchers are actively exploring new technologies. Among them, Reconfigurable Intelligent Surface (RIS), as an innovative wireless communication assistance technology, has attracted much attention. Reconfigurable Intelligent Surface (RIS) is a two-dimensional structure composed of a large number of programmable reflection units, which can dynamically adjust the reflection characteristics of electromagnetic waves, including phase, amplitude, and polarization, through software control. By precisely controlling the parameters of each reflection unit, RIS can achieve beam steering, path optimization, and interference management on the wireless signal propagation path, thereby improving the wireless channel conditions and enhancing the overall performance of the communication system.

[0003] However, since RIS consists of a large number of reflection units and each unit has numerous configuration parameters, how to efficiently design and optimize the overall control strategy to achieve better signal enhancement and interference suppression is a complex optimization problem. Summary of the Invention

[0004] The present invention aims to provide a wireless communication system and method to solve the above technical problems and improve the reliability of wireless communication.

[0005] To solve the above technical problems, the present invention provides a wireless communication system, including:

[0006] A transmitting end, configured to obtain a wireless signal and target channel state information, and perform optimization processing on the target channel state information according to a target optimization model to obtain a RIS configuration instruction, so that the reconfigurable intelligent surface adjusts the reflection units of the reconfigurable intelligent surface based on the RIS configuration instruction, and further enables the reconfigurable intelligent surface to reflect the wireless signal to obtain a reflected signal;

[0007] A user equipment, configured to generate an uplink signal according to the reflected signal and send the uplink signal to the transmitting end.

[0008] In the above solution, the sending end obtains the wireless signal and the target channel state information, where the target channel state information provides the basic data for the optimal configuration of the reconfigurable intelligent surface; the sending end is used to optimize the target channel state information according to the target optimization model, so as to determine better configurations for each reflection unit in the current reconfigurable intelligent surface, generate a more reliable RIS configuration instruction, so that the reconfigurable intelligent surface adjusts the reflection units based on the RIS configuration instruction. Thus, when the reconfigurable intelligent surface reflects the wireless signal transmitted by the sending end, better signal enhancement and interference suppression can be achieved, and a more reliable reflected signal can be obtained to be sent to the user equipment. And the sending end generates RIS configuration instructions that are more in line with the current state in real time based on the target channel state information, meeting the requirements of the dynamically changing wireless environment and improving the reliability of wireless communication.

[0009] Further, the sending end is used to obtain the wireless signal and the target channel state information, including:

[0010] When obtaining the target channel state information, receive the environmental perception data and the direct channel state information sent by the user equipment;

[0011] Based on the direct channel state information and the environmental perception data, construct the RIS channel information through a preset channel inference algorithm;

[0012] Based on the direct channel state information and the RIS channel information, construct the target channel state information.

[0013] In the above solution, through the directly available environmental perception data and direct channel state information data, relatively accurate and difficult-to-obtain RIS-related channel information is obtained based on a preset channel inference algorithm, and then the target-related channel information is obtained, providing the necessary data support for the optimal configuration of the RIS.

[0014] Further, the sending end is used to obtain the wireless signal and the target channel state information, including:

[0015] When obtaining the target channel state information, compress the target channel state information based on the compressive sensing algorithm to obtain low-dimensional information;

[0016] Reconstruct the low-dimensional information based on the signal reconstruction algorithm to obtain the target channel state information.

[0017] Further, when the sending end is used to obtain the target channel state information, compressing the target channel state information based on the compressive sensing model to obtain low-dimensional information includes:

[0018] Obtain a random measurement matrix, where the number of rows of the random measurement matrix is less than the number of rows of the channel matrix of the target channel state information;

[0019] Perform a linear projection on the target channel state information through a random measurement matrix to obtain low-dimensional information.

[0020] In the above solution, by utilizing the sparsity of the signal, the target channel state information is compressed, reducing the acquisition amount and feedback overhead of the target channel state information. At the same time, after obtaining the target channel state information, signal reconstruction is performed to improve the efficiency and accuracy of CSI acquisition.

[0021] Further, the transmitter optimizes the target channel state information according to a target optimization model to obtain a RIS configuration instruction, including:

[0022] Obtain sample configuration data, where the sample state information in the sample configuration data is used as input information, and the sample action information in the sample configuration data is used as output information;

[0023] Based on the sample configuration data, train a preset optimization model through a reward function to obtain the target optimization model;

[0024] Obtain the to-be-measured state information, where the to-be-measured state information includes the target channel state information;

[0025] Process the to-be-measured state information through the target optimization model to obtain target action information;

[0026] Generate the RIS configuration instruction based on the target action information.

[0027] Further, the transmitter is used to train a preset optimization model through a reward function based on the sample configuration data to obtain the target optimization model, including:

[0028] Based on the signal-to-noise ratio algorithm and the sample state information, obtain the signal-to-noise ratio of the user equipment;

[0029] Based on the received power of the interference source at different times, obtain the signal-to-noise ratio of the interference source;

[0030] Construct the reward function based on the signal-to-noise ratio of the user equipment and the signal-to-noise ratio of the interference source.

[0031] In the above solution, the to-be-measured state information is obtained based on the real-time generated target channel state information, thereby realizing the generation of the RIS configuration instruction and dynamically adjusting the parameters of the reflection unit, improving the real-time performance and reliability of the adjustment of the reflection unit in the reconfigurable intelligent surface.

[0032] Further, the transmitter is used to optimize the target channel state information according to the target optimization model, including:

[0033] Predict and complete the target channel state information based on the target prediction and completion model to obtain the complete channel state information;

[0034] Optimize the complete channel state information according to the target optimization model to obtain the RIS configuration instruction.

[0035] Further, the transmitter is used to predict and complete the target channel state information based on the target prediction and completion model to obtain the complete channel state information, including:

[0036] Obtain the sample channel state information;

[0037] Train the preset prediction and completion model through the sample channel state information to obtain the target prediction and completion model;

[0038] Input the target channel state information into the target prediction and completion model to obtain the complete channel state information.

[0039] Further, the transmitter is used to predict and complete the target channel state information based on the target prediction and completion model to obtain the complete channel state information, including:

[0040] Divide the target prediction and completion model into a training set and a test set;

[0041] Train the preset prediction and completion model through the training set to obtain the initial training prediction and completion model;

[0042] Input the test set into the initial training prediction and completion model to obtain the predicted value;

[0043] Evaluate the error of the predicted value to obtain the prediction error. When the prediction error is greater than the preset threshold, adjust the parameters of the initial training prediction and completion model until the prediction error is less than or equal to the preset threshold, and output the target prediction and completion model.

[0044] In the above solution, the target prediction and completion model is used to complete the target channel state information, thus ensuring the accuracy of the subsequent RIS configuration instruction generated using the target channel state information.

[0045] The present invention also provides a wireless communication method, which is applied to a sending end and includes: obtaining a wireless signal and target channel state information; optimizing the target channel state information according to a target optimization model to obtain a RIS configuration instruction, so that a reconfigurable intelligent surface adjusts reflection units of the reconfigurable intelligent surface based on the RIS configuration instruction, and further enables the reconfigurable intelligent surface to reflect the wireless signal to obtain a reflected signal; and then enabling a user equipment to generate an uplink signal according to the reflected signal and send the uplink signal to the sending end.

[0046] The method provided in this embodiment can be executed based on the sending end in the above solution. The sending end obtains a wireless signal and target channel state information, where the target channel state information provides basic data for the optimized configuration of the reconfigurable intelligent surface; the sending end is used to optimize the target channel state information according to the target optimization model, so as to determine better configurations of the respective reflection units in the current reconfigurable intelligent surface, generate a more reliable RIS configuration instruction at present, and enable the reconfigurable intelligent surface to adjust the reflection units based on the RIS configuration instruction, so that when the reconfigurable intelligent surface reflects the wireless signal transmitted by the sending end, better signal enhancement and interference suppression can be achieved, and a more reliable reflected signal can be obtained for sending to the user equipment. Moreover, the sending end generates a RIS configuration instruction that better conforms to the current state in real time based on the target channel state information, meets the requirements of the dynamically changing wireless environment, and improves the reliability of wireless communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of a wireless communication system architecture provided by an embodiment of the present invention;

[0048] Figure 2 It is a schematic flowchart of a wireless communication method provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] Please refer to Figure 1 , this embodiment provides a wireless communication system, including:

[0051] A transmitting end, configured to obtain a wireless signal and target channel state information, and optimize the target channel state information according to a target optimization model to obtain a RIS configuration instruction, so that a reconfigurable intelligent surface adjusts reflection units of the reconfigurable intelligent surface based on the RIS configuration instruction, and further enables the reconfigurable intelligent surface to reflect the wireless signal to obtain a reflected signal;

[0052] A user equipment, configured to generate an uplink signal according to the reflected signal and send the uplink signal to the transmitting end.

[0053] In this embodiment, the transmitting end obtains a wireless signal and target channel state information, where the target channel state information provides basic data for the optimization configuration of the reconfigurable intelligent surface; the transmitting end is configured to optimize the target channel state information according to a target optimization model, so as to determine better configurations of the respective reflection units in the current reconfigurable intelligent surface, generate a more reliable current RIS configuration instruction, and enable the reconfigurable intelligent surface to adjust the reflection units based on the RIS configuration instruction, so that when the reconfigurable intelligent surface reflects the wireless signal transmitted by the transmitting end, better signal enhancement and interference suppression can be achieved, and a more reliable reflected signal can be obtained for sending to the user equipment. Moreover, the transmitting end generates a RIS configuration instruction that is more in line with the current state in real time based on the target channel state information, meets the requirements of the dynamically changing wireless environment, and improves the reliability of wireless communication.

[0054] In another embodiment, the transmitting end is configured to obtain a wireless signal and target channel state information, including:

[0055] Receiving environmental perception data and direct channel state information sent by the user equipment;

[0056] Based on the direct channel state information and the environmental perception data, constructing RIS channel information through a preset channel inference algorithm;

[0057] Based on the direct channel state information and the RIS channel information, constructing the target channel state information.

[0058] It should be noted that the efficient configuration of RIS depends on accurate CSI (Channel State Information). However, in actual deployment, since RIS usually does not have the ability to actively transmit and receive, it is difficult to obtain comprehensive CSI information. This limits the fast response and optimization ability of RIS in a dynamic environment. For the problem that the reconfigurable intelligent surface (RIS) cannot actively transmit and receive signals, resulting in difficulties in obtaining channel state information (CSI), it can be obtained through indirect CSI estimation. Further, the transmitter includes a base station and a control unit. The base station sends known pilot signals to the user equipment. After receiving these signals, the user equipment estimates the direct channel state information between the base station and the user equipment using its own receiving ability. Then, the user equipment transmits the estimated direct channel state information back to the base station through a feedback channel.

[0059] It is expressed as:

[0060] Y BU = H BU X + N;

[0061] H BU is the channel matrix from the Base Station (BS) to the User Equipment (UE), X is the pilot signal vector (known) sent by the base station for channel estimation, Y BU is the signal vector received by the user equipment, that is, the result of the base station pilot passing through the channel and adding noise. N is the additive Gaussian white noise matrix, modeling the noise interference at the receiving end.

[0062] The base station obtains environmental perception data through deployed environmental perception sensors (such as cameras, radars, etc.), including the location of RIS, the location and movement state of the user equipment, the location of environmental obstacles, etc. Combine these environmental perception data with the direct channel state information H BU received by the base station, and use the geometric channel model to infer the RIS-related RIS channel H BR and H RU .

[0063] The geometric channel model is expressed as:

[0064]

[0065] where, H BR is the channel between the base station and RIS, H RU is the channel between RIS and the user equipment, α l and β m are path gains, a R (θ l ), a B(φ l )、a U (θ m ) and a R (φ m ) are the antenna array direction vectors of the RIS, the base station, and the user equipment respectively. Since the two vectors of the RIS correspond to the receiving and transmitting roles respectively, a total of four array response vectors are required. Specifically, a R (θ l ) is the array response vector of the RIS as the "receiver" for the incident wave from the base station in the l-th multipath; a B (φ l ): is the array response vector of the base station as the "transmitter" for the wave radiated in the direction of the RIS in the l-th multipath; a U (θ m ) is the array response vector of the user equipment as the "receiver" for the arriving wave from the RIS in the m-th multipath; a R (φ m ) is the array response vector of the RIS as the "transmitter" for the wave radiated in the direction of the user equipment in the m-th multipath. L and M are the number of multipaths. In complex baseband wireless channel modeling, the array response of the transmitter is usually denoted by the superscript H to represent the conjugate transpose, so as to form an inner product with the receiver response. Therefore, the direction vector of the RIS in the link should also have the superscript H. Thus, the conjugate transpose of a B (φ l ) and the conjugate transpose of a R (φ m )

[0066] are obtained. Then, using the Bayesian estimation or least squares estimation method, the RIS-related RIS channel state information is inferred. Specifically as follows:

[0067]

[0068] where, ||·|| F represents the Frobenius norm, and λ is the regularization parameter, which is used to control the stability and accuracy of the estimation.

[0069] Through the indirect CSI estimation method, relatively accurate RIS-related channel information can be obtained without increasing the hardware complexity of the RIS, providing the necessary data support for the optimal configuration of the RIS.

[0070] In another embodiment, the transmitter is used to obtain the wireless signal and the target channel state information, including:

[0071] When obtaining the target channel state information, the target channel state information is compressed based on the compressed sensing algorithm to obtain low-dimensional information;

[0072] Reconstruct the low-dimensional information based on the signal reconstruction algorithm to obtain the target channel state information.

[0073] It should be noted that compressive sensing is a theoretical method based on signal sparsity, which can recover the original signal through a small amount of sampling data. In wireless communication, CSI often has sparsity, which means that most elements of the channel matrix H are zero or close to zero, or can be made sparse in a certain domain through appropriate basis transformation. Therefore, compressive sensing can use fewer measurement data to recover the high-dimensional channel state information. In practical applications, the channel matrix H of the RIS channel state information in the RIS system is high-dimensional, and traditional acquisition methods require a large number of channel measurements, which will bring high overhead. Compressive sensing technology reduces the required number of sampling points by using the sparsity principle during the measurement process, thereby reducing the burden of data acquisition and transmission. Therefore, within the transmitter, sometimes it is necessary to transmit the channel state information. For example, when the base station at the transmitter obtains the RIS channel state information, it sends the RIS channel state information and the direct channel state information as the target channel state information to the control unit in the transmitter. At this time, it is necessary to compress the target channel state information based on the compressive sensing algorithm to obtain low-dimensional information. After that, after the acquisition is completed, it is necessary to reconstruct the low-dimensional information. Exemplarily, the control unit in the transmitter reconstructs it after receiving the low-dimensional information to obtain the target channel state information. Among them, signal reconstruction mainly uses signal reconstruction algorithms, such as matching pursuit, basis pursuit algorithm, orthogonal matching pursuit, etc., to reconstruct the low-dimensional information and recover the high-dimensional target channel state information.

[0074] Specifically, the channel sparse representation is as follows: In high-frequency band (such as millimeter wave and terahertz wave) communication, the channel usually has sparse characteristics, that is, there are fewer multipath propagation paths. Using sparse representation, the high-dimensional CSI can be represented as a low-dimensional sparse signal, thereby reducing the amount of data collected and transmitted. The high-dimensional target channel state information H is represented as:

[0075] H = ΨΘ;

[0076] where Ψ is the sparse basis and Θ is the sparse coefficient matrix.

[0077] The compressive sensing algorithm is as follows: Obtain a random measurement matrix, where the number of rows of the random measurement matrix is less than the number of rows of the channel matrix of the target channel state information; perform a linear projection on the target channel state information through the random measurement matrix to obtain low-dimensional information. Specifically, use the random measurement matrix Φ to perform a linear projection on the high-dimensional CSI, that is, the target channel state information, to obtain the low-dimensional information y to reduce the amount of feedback data.

[0078] y = ΦH;

[0079] Among them, y is low-dimensional information. is a random measurement matrix, where M′ << M. M is the dimension of the original channel vector; M′ is the number of compressed measurements.

[0080] The signal reconstruction algorithm is as follows:

[0081] Using signal reconstruction algorithms such as matching pursuit, basis pursuit algorithm, orthogonal matching pursuit, etc., the low-dimensional information y is reconstructed to recover the high-dimensional target channel state information H.

[0082] The reconstruction objective is:

[0083]

[0084] Among them, ||||1 is the L1 norm, which is used to promote sparse solutions, ∈ is the error tolerance. ||||2 is the L2 norm, and subject to ensures that the reconstructed low-dimensional observations are consistent with the actual measurements - within the noise tolerance range.

[0085] Through the compressive sensing algorithm, only low-dimensional measurement data needs to be transmitted, greatly reducing the feedback overhead of CSI. At the same time, the signal reconstruction algorithm guarantees the accurate recovery of CSI, ensuring the reliability of subsequent generation of RIS configuration instructions. Applying compressive sensing technology can significantly reduce the feedback data volume and computational overhead while ensuring the accuracy of CSI recovery, improving the overall efficiency and real-time performance of the system.

[0086] Furthermore, multiple control units can be set in the sending end to achieve distributed CSI management. The distributed CSI management method distributes the tasks of CSI acquisition, processing, and management to multiple control units, reducing the computational burden of a single control unit while improving the real-time performance and reliability of CSI acquisition. The specific implementation steps are as follows:

[0087] (1) Control unit distribution:

[0088] Multiple control units are deployed in the system, and each control unit is responsible for managing the CSI acquisition and processing of a specific area or a specific RIS module. In this way, the CSI management tasks of the entire system are dispersed to multiple control nodes, avoiding single-point overload.

[0089] (2) Local CSI processing:

[0090] Each control unit independently processes the CSI information in its responsible area, including steps such as channel estimation, sparse reconstruction, and channel fusion. Local processing can reduce the data transmission volume and processing delay, improving real-time performance.

[0091] (3) Global CSI fusion:

[0092] Each control unit fuses the locally processed CSI information through a high-speed network connection to form a global CSI matrix. The global CSI is used for the global optimization of RIS configuration to ensure the synergistic effect of signal enhancement and interference management, expressed as:

[0093]

[0094] where C is the number of control units, and H c is the local CSI matrix processed by the Cth control unit.

[0095] Through a dynamic load balancing strategy, the CSI management tasks are reasonably allocated to different control units to avoid overloading some control units. At the same time, a fault tolerance mechanism is designed so that when some control units fail, other control units can take over their tasks to ensure the stable operation of the system. Using a distributed computing framework, it supports parallel processing of multiple CSI management tasks, speeds up the CSI acquisition and processing speed, and improves the real-time response ability of the system. Through distributed CSI management, the system can maintain high-efficiency CSI acquisition and management capabilities in large-scale RIS deployment and high-density user scenarios, and improve the real-time performance and accuracy of RIS configuration.

[0096] In another embodiment, the sender optimizes the target channel state information according to the target optimization model to obtain an RIS configuration instruction, including:

[0097] Obtain sample configuration data, where the sample state information in the sample configuration data is used as input information, and the sample action information in the sample configuration data is used as output information;

[0098] Based on the sample configuration data, train a preset optimization model through a reward function to obtain the target optimization model;

[0099] Obtain the to-be-measured state information, where the to-be-measured state information includes the target channel state information;

[0100] Process the to-be-measured state information through the target optimization model to obtain the target action information;

[0101] Generate an RIS configuration instruction based on the target action information.

[0102] It should be noted that when obtaining the sample configuration data, the sample state information in the sample configuration data is used as input information, where the state information is used to describe the information of the current environment, including:

[0103] (1) Channel state information (CSI):

[0104] The channel between the base station and the RIS: denoted as where M is the number of base station antennas and N is the number of RIS reflection units.

[0105] The channel between the RIS and the user equipment: denoted as where K is the number of user equipments.

[0106] (2) User equipment location and movement state:

[0107] The location vector of the user equipment

[0108] The movement speed and direction of the user equipment.

[0109] (3) Environmental interference source information:

[0110] The location and channel state of the interference source.

[0111] (4) Current RIS configuration:

[0112] The phase and amplitude settings of each reflection unit of the RIS

[0113] Combining the above information, the state s t can be expressed as:

[0114] s t =(H BR , H RU , {p k}, {v k}, Θ);

[0115] where, represents the set of RIS reflection units, n is the serial number of the reflection unit, and v k is the speed vector of the k-th user equipment.

[0116] Secondly, the sample action information in the sample configuration data is used as the output information. The action information refers to the operations that the intelligent agent can take in the current state. In this application, the action information involves the phase and amplitude adjustment of each reflection unit of the RIS. The specific design is as follows:

[0117] (1) Phase adjustment:

[0118] The phase θ of each reflection unit n can be discretized into L discrete values, that is:

[0119] θ n ∈ {0, Δθ, 2Δθ, …, (L - 1)Δθ}, where

[0120] (2) Amplitude adjustment:

[0121] Assume that the amplitude adjustment is fixed at 1 (total reflection), then only the phase needs to be adjusted; if the amplitude needs to be adjusted, the same discretization process can also be carried out.

[0122] Therefore, the action information a t at time step t can be expressed as:

[0123]

[0124] represents the index set of all programmable reflection units (unit cells) on the RIS. n is the serial number of the reflection unit. The RIS consists of mutually independent reflection units (also called meta-atoms or unit cells), and these units are regularly arranged on a two-dimensional plane. Each reflection unit can independently adjust the reflection phase (and optionally the amplitude) of the electromagnetic wave, is the total number of units of the RIS, which determines the degree of freedom and spatial resolution of the surface for the electromagnetic wave.

[0125] To reduce the dimension of the action space, a hierarchical strategy can be adopted. For example, the reflection units are grouped, and each group adjusts the phase uniformly, thereby reducing the number of parameters that need to be decided.

[0126] After that, based on the sample configuration data, a preset optimization model is trained through a reward function to obtain a target optimization model. Specifically, the Deep Q-Network (DQN) and Deep Deterministic Policy Gradient (DDPG) in the deep reinforcement learning algorithm are used for model training to optimize the RIS configuration. The specific steps are as follows:

[0127] (1) Initialization: Initialize the parameters of the deep Q-network or DDPG model; initialize the experience replay buffer D.

[0128] (2) Experience collection: At each time step t, based on the current state information s t , an action information a t is selected through an ε-greedy policy. After executing the action information, the reward R t in the reward function and the next state information s t+1 are observed; the experience tuple (s t , a t , R t , s t+1 ) is stored in the experience replay buffer D.

[0129] (3) Batch training: Randomly sample a batch of experiences from the experience replay buffer for training the preset optimization model. For DQN, update the Q-network parameters to minimize the following loss function:

[0130]

[0131] Among them, s, a, r, s' are the current state information, action information, reward result in the reward function, and next state information respectively, γ is the discount factor, and θ - are the target network parameters. Q() calculates the parameters by the deep Q network.

[0132] For DDPG, update the parameters of the policy network and the value network to maximize the expected cumulative reward.

[0133] (4) Policy update:

[0134] Regularly synchronize the online network parameters to the target network to ensure the stability of training.

[0135] Use the gradient descent method to optimize the network parameters and improve the decision-making ability of the model.

[0136] (5) Training iteration:

[0137] Repeat the above process until the model converges, that is, in most states, the selected action can maximize the cumulative reward to obtain the target optimization model.

[0138] After the training is completed, deploy the optimized target optimization model to the control unit in the transmitter to achieve real-time optimization of RIS configuration. Specifically: Model deployment: Import the trained target optimization model into the processor in the control unit; Configure the model running environment to ensure that the target optimization model can efficiently process the real-time state information to be measured and output RIS configuration instructions; The control unit receives the target channel state information including CSI channel state information transmitted by the base station through the feedback channel in real time and updates the current state information s t ; Based on the current state information s t , the target optimization model calculates and generates the target action information a t , that is, the phase and amplitude adjustment parameters of each reflection unit of RIS, and generates RIS configuration instructions based on the target action information. Send the generated RIS configuration instructions to the Reconfigurable Intelligent Surface (RIS) through the feedback channel to achieve dynamic adjustment of the reflection parameters. The control unit in the transmitter continuously monitors the changes in the wireless environment. By continuously receiving new target channel state information, the target optimization model adjusts the RIS configuration in real time to ensure continuous optimization of signal enhancement and interference suppression. If it is detected that the environment has changed significantly (such as user movement, change of interference source), the target model can respond quickly and adjust the RIS configuration to meet the new environmental requirements.

[0139] Further, the transmitter is used to train a preset optimization model based on sample configuration data through a reward function to obtain a target optimization model, including:

[0140] Based on the signal-to-noise ratio algorithm and sample status information, obtain the signal-to-noise ratio of the user equipment;

[0141] Based on the received power of the interference source at different times, obtain the signal-to-noise ratio of the interference source;

[0142] Based on the signal-to-noise ratio of the user equipment and the signal-to-noise ratio of the interference source, construct a reward function.

[0143] It should be noted that the reward function is used to evaluate the effect of the current action and guide the agent to learn an optimization strategy. In this application, it is specifically constructed as follows:

[0144] The received signal-to-noise ratio SNR of the target device target , the improvement compared to the t-1 moment at time t is:

[0145]

[0146] The reduction in the received power of the interference signal of the interference source is:

[0147]

[0148] Among them, ΔInterference is the reduction in the received power of the interference signal, is the set of interference sources, P i is the received power of the i-th interference source.

[0149] The constructed reward function R t can be expressed as:

[0150] R t =α·ΔSNR target -β·ΔInterference;

[0151] Among them, α and β are weight coefficients used to balance the importance of signal enhancement and interference suppression.

[0152] Further, to better describe the model for RIS configuration optimization, the following symbols and formulas are set:

[0153] Base station transmitted signal After passing through the channel H between the base station and the RIS BR , the RIS configuration matrix Θ = diag(e j θ 1,e jθ 2,…,e jθ N), and the channel H between the RIS and the user equipment RU , the signal y finally reaching the user equipment can be expressed as:

[0154] y = H RU ΘH BR x + n;

[0155] where n is the received noise.

[0156] In the calculation of the signal-to-noise ratio (SNR), the SNR of the target user equipment target can be expressed as:

[0157]

[0158] where h RU,k and h BR,k are the channel vectors between the k-th user equipment and the RIS, and between the base station and the RIS respectively, and σ 2 is the noise power, and the base station transmits the signal

[0159] The optimization objective can be set as: The goal of optimization is to maximize the SNR of the target user equipment while minimizing the SNR of the interference source of the interfering user equipment. Specifically, it can be expressed as:

[0160]

[0161] where, and represent the sets of the target user equipment and the interfering user equipment respectively, and λ is a weight coefficient used to balance the importance of signal enhancement and interference suppression. Search for the optimal solution under all possible RIS phase settings, and perform value search on the configuration parameter Θ of the RIS to maximize the overall performance index of the system.

[0162] Thus, learn the optimal RIS configuration strategy to achieve dynamic enhancement of wireless signals and effective management of interference.

[0163] In another embodiment, the sending end is used to optimize the target channel state information according to the target optimization model, including:

[0164] Predict and complete the target channel state information based on the target prediction completion model to obtain the complete channel state information;

[0165] Optimize the complete channel state information according to the target optimization model to obtain the RIS configuration instruction.

[0166] In another embodiment, the sending end is used to predict and complete the target channel state information based on the target prediction completion model to obtain the complete channel state information, including:

[0167] Obtain sample channel state information;

[0168] Train a preset prediction completion model with the sample channel state information to obtain a target prediction completion model;

[0169] Input the target channel state information into the target prediction completion model to obtain the complete channel state information.

[0170] In another embodiment, the transmitter is used to predict and complete the target channel state information based on the target prediction completion model to obtain the complete channel state information, including:

[0171] Divide the historical channel state information into a training set and a test set;

[0172] Train a preset prediction completion model with the training set to obtain an initial training prediction completion model;

[0173] Input the test set into the initial training prediction completion model to obtain predicted values;

[0174] Evaluate the error of the predicted values to obtain a prediction error. When the prediction error is greater than a preset threshold, adjust the parameters of the initial training prediction completion model until the prediction error is less than or equal to the preset threshold, and output the target prediction completion model.

[0175] It should be noted that the target prediction completion model uses time series prediction technology in machine learning to predict and complete the missing or delayed CSI information to ensure the accuracy and timeliness of the RIS configuration. The specific implementation steps are as follows:

[0176] (1) Construct a preset prediction completion model:

[0177] Regard the target channel state information as a time series signal, and use machine learning models such as Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), or Transformer model to establish a preset prediction completion model for CSI.

[0178] (2) Model training:

[0179] Collect historical channel state information, where the historical channel state information is the historical information of the target channel state information, and train the preset prediction completion model according to the historical channel state information so that it can learn the dynamic change law of the target channel state information. The training objective is to minimize the error between the predicted target channel state information and the actual channel state information. The loss function is defined as:

[0180]

[0181] Among them, is the predicted target channel state information, H(t) is the actual target channel state information, and T is the number of training samples. When the loss function converges, the test set is input into the initial training prediction completion model to obtain the predicted value. The predicted value is subjected to error control and confidence evaluation. When the prediction error exceeds the preset threshold, re-estimation or adjustment of the prediction model is triggered to ensure the reliability of the RIS configuration.

[0182] (3) CSI Prediction:

[0183] Using the trained target prediction completion model, the target channel state information for future time steps is predicted to fill in the missing target channel state information caused by channel estimation delay or feedback loss. It is expressed as:

[0184]

[0185] Among them, f is the prediction model, N is the number of historical time steps, Δt is the prediction step length, and H(t) is the target channel state information at time t.

[0186] When the actual target channel state information is partially missing or arrives late, the target channel state information generated by the target prediction completion model is used for completion to obtain the complete channel state information, ensuring that the RIS configuration algorithm can be optimized based on the complete and accurate CSI. Through the online learning and model update mechanism, the parameters of the prediction model are dynamically adjusted to adapt to the changes in the wireless environment, improving the prediction accuracy and the robustness of the model. The model update rule can adopt the Incremental Learning strategy, enabling the model to quickly adjust when new data arrives without having to retrain the entire model. Through the CSI prediction and completion method, the uncertainty and delay problems in the CSI acquisition process can be effectively addressed, ensuring the accuracy and real-time performance of the RIS configuration and improving the overall performance of the wireless communication system.

[0187] Furthermore, in order to reduce the manufacturing cost of the Reconfigurable Intelligent Surface (RIS) and improve its control accuracy, multiple optimizations can be carried out in the hardware design aspect. Specific measures include modular reflection unit design, low-power circuit design, development of integrated control chips, and optimization of manufacturing processes. Specifically,

[0188] (1) Modular Reflection Unit Design

[0189] The modular reflective unit design simplifies the system and reduces costs by dividing the RIS reflective unit into multiple independently controllable small modules, while improving the system's scalability and maintenance convenience.

[0190] Modular structure:

[0191] The overall structure of the RIS is divided into several modules, and each module contains several reflective units. This design enables the system to flexibly expand or reduce the number of modules according to needs, adapting to application scenarios of different scales. For example, if a module contains 16 reflective units, a RIS with 64 reflective units can be composed of 4 modules.

[0192] Independent control:

[0193] Each module is equipped with an independent controller that can individually adjust the phase and amplitude of the internal reflective units. This not only simplifies the design of the overall control strategy but also improves the system's fault tolerance. If a certain module fails, the other modules can still operate normally, ensuring the stability and reliability of the system.

[0194] Cost advantage:

[0195] The modular design reduces manufacturing costs by mass-producing standardized modules. At the same time, the modular structure is convenient for maintenance and upgrading, reducing the overall operating cost of the system.

[0196] Scalability:

[0197] The modular design enables the RIS system to flexibly expand according to actual needs. For example, in areas with a dense user population, the number of modules can be increased to improve signal coverage and enhancement effects; in areas with a sparse user population, the number of modules can be reduced to save costs.

[0198] Let the total number of reflective units of the RIS be N, the number of reflective units contained in each module be n, then the required number of modules M is:

[0199]

[0200] where, denotes rounding up.

[0201] (2) Low-power circuit design

[0202] The low-power circuit design aims to optimize the power consumption management of the RIS reflective unit to ensure the overall energy efficiency of the system during large-scale deployment.

[0203] Power consumption management strategy:

[0204] Adopt a variety of power management techniques, such as Dynamic Voltage Scaling (DVS), Power Gating, and Clock Gating, to dynamically adjust the power supply according to the working state of the reflection unit and reduce unnecessary energy consumption.

[0205] High-efficiency circuit components:

[0206] Select low-power and high-performance circuit components, such as low-power Operational Amplifiers (Op-Amps) and high-performance Power Converters, to reduce the overall system power consumption.

[0207] Energy harvesting and management:

[0208] Integrate energy harvesting technologies, such as solar panels or Wireless Power Transfer (WPT), to provide auxiliary power for the RIS and further reduce the dependence on traditional power supplies.

[0209] Thermal optimization:

[0210] Adopt high-efficiency heat dissipation materials and structural designs to ensure that low-power circuits maintain good heat dissipation performance under high-density deployment conditions and prevent performance degradation or hardware damage caused by overheating.

[0211] (3) Integrated control chip

[0212] The development of the integrated control chip aims to integrate the control logic of multiple reflection units into a single chip, reduce the hardware volume and cost, and improve the control accuracy and response speed at the same time.

[0213] System-on-Chip (SoC):

[0214] Develop an integrated chip containing a processor, memory, and control logic to uniformly manage the reflection units of the RIS. The SoC design can significantly reduce the number of external components, reduce system complexity and manufacturing costs.

[0215] High-precision control:

[0216] The integrated chip is built with high-precision Digital-to-Analog Converters (DACs) and Analog-to-Digital Converters (ADCs) to achieve precise control of the phase and amplitude of the reflection unit.

[0217] Low-latency communication interface:

[0218] Design high-speed communication interfaces, such as SPI (Serial Peripheral Interface), I 2 2C (Inter-Integrated Circuit), to ensure that control instructions can be transmitted to each reflection unit quickly and with low latency, improving the system's response speed and real-time performance.

[0219] Programmability and flexibility:

[0220] Integrate programmable logic units (PLUs) and programmable memories, enabling the control chip to be flexibly configured and upgraded according to different application requirements, extending the system's service life.

[0221] Assume that the resources required for the control logic of each reflection unit are R u , and the total resources of the integrated control chip are R chip Then it satisfies: R chip ≥N×R u , where N is the number of reflection units.

[0222] By optimizing the integrated design, the resource utilization rate can be maximized: Maximize η, minimize the chip area A and power consumption P chip : minimize A, P chip , specifically:

[0223]

[0224] Among them, N: the total number of reflection units (unit cells) on the RIS. It determines the degrees of freedom and spatial resolution of surface adjustment. R u : the "amount of resources" required for a single reflection unit on the control chip, including all hardware resources. R chip : the total amount of available resources on the control chip (SoC), also measured in logic units, memory units, or other unified hardware resources. η: "resource utilization rate", indicating the maximum number of reflection units that can be supported under a given chip resource budget. The larger the value, the more RIS units can be served with the same chip resources. A: the silicon chip area of the control chip (chip area), usually in square millimeters (mm 2 2). The smaller the chip area, the lower the manufacturing cost and power consumption tend to be. P chip : the power consumption of the control chip (power consumption), usually in watts (W) or milliwatts (mW). In mobile or energy-constrained scenarios, the lower the chip power consumption, the more beneficial it is to the overall energy efficiency of the system.

[0225] (4) Optimization of manufacturing process

[0226] The optimization of the manufacturing process improves the consistency and reliability of the RIS reflection units and reduces the manufacturing cost by adopting advanced Printed Circuit Board (PCB) technology and microelectronics manufacturing processes.

[0227] Advanced PCB technology:

[0228] Adopt High-Density Interconnect (HDI) PCB technology to achieve a compact layout and efficient connection of the reflection units. HDI PCB can support higher line density and smaller component size, improving the integration and performance of the system.

[0229] Microelectronics manufacturing process:

[0230] Introduce advanced microelectronics manufacturing processes, such as Miniaturization and Surface Mount Technology (SMT), to improve the manufacturing precision and consistency of the reflection units. Precise manufacturing processes can reduce the performance differences between reflection units and improve the overall signal consistency and reliability of the system.

[0231] Automated production line:

[0232] Establish an automated production line and use robots and automated equipment for the assembly and testing of reflection units to improve production efficiency and reduce labor costs. At the same time, automated production can ensure the manufacturing quality and consistency of each reflection unit and reduce the defect rate.

[0233] Material optimization:

[0234] Select high-performance and low-cost materials, such as highly conductive copper foil and low-dielectric-constant substrate materials, to improve the electromagnetic performance and durability of the reflection units. At the same time, optimizing material selection can reduce the overall manufacturing cost and improve the economy of the system.

[0235] Through hardware design and cost optimization measures, modular design and manufacturing process optimization have effectively reduced the manufacturing cost and operating cost of RIS. The integrated control chip reduces the number of hardware components, further reducing the overall system cost; the integrated control chip and high-precision low-power circuit design improve the control precision of the RIS reflection units, ensuring the effectiveness and stability of signal enhancement and interference suppression;

[0236] The modular design and distributed control mechanism enhance the fault tolerance and reliability of the system. Even if some modules or reflection units malfunction, the overall performance of the system can still remain stable; the low-power circuit design and efficient energy management strategy significantly improve the overall energy efficiency of the system, reduce the energy consumption requirements during large-scale deployment, and conform to the development trend of green communication; the modular design and integrated control chip endow the RIS system with high scalability and flexibility, enabling it to be flexibly adjusted and expanded according to actual needs and adapt to wireless communication environments of different scales and complexities. In summary, through the modular reflection unit design, low-power circuit design, development of integrated control chips, and optimization of manufacturing processes, the manufacturing cost of RIS is effectively reduced, the control accuracy and system reliability are improved, the energy efficiency and scalability of the system are optimized, providing a solid hardware foundation for the wide application of RIS in wireless communication systems.

[0237] Furthermore, aiming at the deficiencies of existing algorithms in terms of real-time performance and stability, this application proposes an efficient real-time optimization algorithm, aiming to improve the performance and response speed of deep reinforcement learning (DRL) in RIS configuration optimization through various technical means. Specifically, it includes four aspects: parallel computing architecture, incremental learning strategy, model compression and acceleration, and hybrid optimization algorithm. The implementation methods of these technical means and their roles in the optimization process will be elaborated in detail below.

[0238] (1) Parallel computing architecture

[0239] The parallel computing architecture utilizes high-performance parallel computing resources such as GPUs (Graphics Processing Units) or FPGAs (Field Programmable Gate Arrays) to accelerate the training and inference processes of deep reinforcement learning models, thus significantly improving the real-time performance of the algorithm. The specific implementation steps are as follows:

[0240] GPU acceleration:

[0241] Parallel processing ability: GPUs have the ability of large-scale parallel processing, can handle a large number of computing tasks simultaneously, and are particularly suitable for matrix operations and the forward and backward propagation processes of neural networks.

[0242] Support for deep learning frameworks: Mainstream deep learning frameworks (such as TensorFlow, PyTorch) all support GPU acceleration and achieve efficient utilization of computing resources through technologies such as CUDA.

[0243] GPUs accelerate the gradient calculation and weight update processes through parallel computing.

[0244] FPGA acceleration:

[0245] Customized Computing: FPGAs can customize the hardware logic according to specific requirements to achieve efficient execution of specific computing tasks, such as convolution operations and activation function calculations.

[0246] Low Latency Response: FPGAs have low latency characteristics and are suitable for real-time optimization scenarios that require fast response.

[0247] Through the application of a parallel computing architecture, the training time and inference latency of deep reinforcement learning models are significantly reduced, meeting the requirements of real-time optimization. For example, under GPU acceleration, the model training speed is increased by about 5 times; under FPGA acceleration, the inference latency is reduced to the millisecond level, significantly improving the real-time response ability of the system.

[0248] (2) Incremental Learning Strategy

[0249] The incremental learning strategy aims to enable the model to quickly update when new data arrives without having to retrain from scratch, thereby reducing optimization latency and improving the adaptability of the system. The specific implementation steps are as follows:

[0250] Online Learning: Data Stream Processing: Adopt online learning algorithms to enable the model to gradually learn new data while retaining existing knowledge; Parameter Update: Update the model parameters in real time through methods such as Mini-batch Gradient Descent or Stochastic Gradient Descent (SGD).

[0251] The incremental learning strategy enables the model to quickly adapt to the dynamically changing wireless environment, reducing the time cost of retraining. For example, the model parameter update speed is increased while maintaining a performance level similar to batch training, ensuring the real-time optimization of RIS configuration.

[0252] (3) Model Compression and Acceleration

[0253] Model compression and acceleration speed up the inference speed of the model by reducing the number of parameters and the amount of computation of the deep reinforcement learning model, further improving the real-time performance of the algorithm. The specific implementation steps are as follows:

[0254] Model Pruning: Weight Pruning: Remove neuron connections with small weight values or little impact on the model output to reduce the complexity of the model; Structural Pruning: Delete entire neurons or channels to further compress the model size.

[0255] Quantization: Weight Quantization: Convert weight values from high precision (such as 32-bit floating-point numbers) to low precision (such as 8-bit integers) to reduce storage and computation overhead; Activation Quantization: Similarly quantize activation values to further reduce computational complexity.

[0256] Through model compression and acceleration techniques, the number of parameters of the deep reinforcement learning model is reduced, the inference speed is increased, and the loss of model performance is within 5%. This enables the RIS configuration optimization algorithm to operate efficiently in resource-constrained environments and meet real-time optimization requirements.

[0257] (4) Hybrid optimization algorithm

[0258] The hybrid optimization algorithm combines traditional optimization algorithms (such as gradient descent, heuristic algorithms) with deep reinforcement learning algorithms, making full use of their respective advantages to improve the efficiency and stability of the optimization process. The specific implementation steps are as follows:

[0259] Combination of gradient descent and deep reinforcement learning: Initial optimization: Use the gradient descent algorithm to preliminarily optimize the model parameters and quickly approach the global optimal solution; Reinforcement learning fine-tuning: On the basis of gradient descent optimization, use the deep reinforcement learning algorithm for fine-tuning to further improve the decision-making ability and optimization accuracy of the model. The optimization process can be expressed as:

[0260]

[0261] where, W t : represents the set of all trainable parameters (weights and biases) of the neural network at the "t-th step". W t+1 : represents the new parameter set after a parameter update (i.e., after applying the current gradient descent and reinforcement learning fine-tuning on the basis of the t-th step). η: learning rate, a positive scalar hyperparameter used to control the step size of the ordinary gradient descent term The larger the learning rate, the larger the steps of gradient descent and the faster the convergence, but it may be unstable; the smaller the learning rate, the smoother the convergence but the slower the speed. represents the gradient (i.e., the partial derivative vector) of the loss function L(·) with respect to the current parameter W t . It tells us "how the loss will change if the parameters are adjusted along the gradient direction". Multiplying it by η and taking the negative sign forms the classic update amount of gradient descent. L(W t ) is the loss function (error, negative expected return of policy gradient, cross-entropy loss, etc.) used in this training or optimization. It measures the performance of the current model in a given sample or environment interaction. ΔW RL : represents the additional parameter update amount brought by the policy fine-tuning in the Deep Reinforcement Learning algorithm.

[0262] Combination of Heuristic Algorithm and Deep Reinforcement Learning: Heuristic Search: Use heuristic algorithms (such as genetic algorithms, particle swarm optimization) to perform global search in a larger area to find the initial value of a better RIS configuration; Reinforcement Learning for Local Optimization: Based on the initial value found by the heuristic algorithm, use the deep reinforcement learning algorithm for local optimization to improve the accuracy and stability of the configuration.

[0263] The initial configuration Θ0 provided by the heuristic search is optimized through reinforcement learning:

[0264] Θ final = RL_optimize(Θ0, s t );

[0265] where Θ final : The finally output RIS configuration matrix (or phase vector), which contains the optimized phase / amplitude settings of all N reflecting elements at time step t for actual electromagnetic wave reflection. RL_optimize: A "deep reinforcement learning optimization" function / process. Based on the input initial configuration and the current environmental state, after several steps of policy evaluation and update, it outputs the optimal or approximately optimal RIS configuration. It may internally contain typical RL operations such as sampling, neural network forward / backward propagation, action selection, experience replay, etc. Θ0: The "initial RIS configuration" of the optimization process, the optimal configuration of the previous time step. s t : The "environmental state" vector / information set at time step t, usually including: real-time channel state information (CSI), environmental parameters such as the location and speed of the user equipment; information about existing interference sources.

[0266] Joint Optimization Strategy: Multi-Task Learning: Jointly train traditional optimization tasks and reinforcement learning tasks so that the model can optimize both the signal enhancement and interference suppression objectives simultaneously; Strategy Fusion: Design a fusion strategy, use the output of traditional optimization algorithms as the input of reinforcement learning algorithms to form a collaborative optimization mechanism and improve the overall optimization effect. The joint optimization objective function can be expressed as:

[0267]

[0268] where α, β, and γ are weight coefficients, and Heuristic Score is the evaluation score of the heuristic algorithm.

[0269] The hybrid optimization algorithm significantly improves the efficiency and stability of RIS configuration optimization by combining the advantages of traditional optimization methods and deep reinforcement learning. The convergence speed in the optimization process is increased, and the adaptability of the system in different environments is significantly enhanced, ensuring the efficiency and reliability of the RIS configuration.

[0270] Furthermore, to address the deficiencies and challenges in the integration and standardization of Reconfigurable Intelligent Surface (RIS) with existing communication systems, this application proposes a series of system integration and standardization methods. These methods aim to ensure that RIS technology can seamlessly integrate into the existing wireless communication infrastructure, improve system compatibility and interoperability, and promote the wide application and popularization of RIS technology. Specifically, it includes aspects such as unified interface design, protocol adaptation layer, modular system architecture, standardized optimization algorithms, and cross-layer collaborative optimization.

[0271] (1) Unified interface design

[0272] The unified interface design aims to formulate unified communication interface standards between RIS and Base Station (BS), User Equipment (UE), ensuring compatibility and interoperability between different manufacturers and devices.

[0273] A. Interface standard formulation:

[0274] Communication interface protocol: Formulate a unified communication protocol to define the data transmission format, control signaling, and communication process between RIS and the base station. This protocol needs to be compatible with existing wireless communication standards (such as 5G NR) to ensure that RIS can operate stably in different network environments.

[0275] Physical layer interface: Define the physical connection standards between RIS and the base station, including antenna interfaces, frequency band support, and power control, etc., to ensure the efficiency and stability of signal transmission.

[0276] B. Modular interface design:

[0277] Pluggable interface module: Design a pluggable interface module to enable RIS to easily physically connect to base stations and user equipment from different manufacturers, simplifying the installation and maintenance process.

[0278] Standardized interface documentation: Compile detailed interface documentation covering communication protocols, data formats, signaling processes, etc., to guide different device manufacturers to achieve compatibility.

[0279] C. Interface compatibility testing:

[0280] Interoperability testing: Establish an interoperability testing platform to conduct compatibility tests on RIS and base stations, user equipment from different manufacturers and models to ensure the effectiveness of the unified interface design.

[0281] Authentication mechanism: Introduce an authentication mechanism to authenticate RIS devices that meet the unified interface standard to ensure their compatibility and reliability in actual deployment.

[0282] The signal transmission model in the unified interface design can be expressed as:

[0283] Y = H BS-RIS X BS + N;

[0284] Where: Y is the signal received by the RIS; H BS-RIS is the channel matrix between the base station and the RIS; X BS is the signal sent by the base station; N is the noise matrix. Through the standardized interface protocol, ensure that the channel matrix H in the signal transmission process BS-RIS can be accurately estimated and optimized, thereby improving the overall performance of the system.

[0285] (2) Protocol Adaptation Layer

[0286] The protocol adaptation layer introduces a protocol adaptation layer in the RIS system, maps and converts the control instructions of the RIS with existing communication protocols (such as 5G NR protocol), and realizes seamless integration.

[0287] A. Protocol Conversion Mechanism:

[0288] Control Signaling Conversion: Develop a protocol conversion module to convert the control instructions of the RIS (such as reflection phase adjustment, amplitude adjustment) into a signaling format compatible with the existing communication protocol. For example, convert the phase adjustment instruction of the RIS into the control message format of 5G NR to ensure that the instruction can be correctly parsed and executed by the base station.

[0289] Two-way Communication Bridging: Establish a two-way communication bridge between the RIS and the base station to ensure that control instructions and status information can be transmitted efficiently and reliably between the RIS and the base station.

[0290] B. Protocol Adaptation Layer Architecture:

[0291] Middleware Design: Design protocol adaptation middleware as a bridge between the RIS system and the existing communication protocol, responsible for tasks such as signaling conversion, data format matching, and protocol synchronization.

[0292] Scalability: The protocol adaptation layer needs to have good scalability to support the upgrade and expansion of future communication protocols and ensure the long-term compatibility of the RIS system.

[0293] C. Protocol Adaptation Optimization:

[0294] Latency Minimization: Optimize the processing flow of the protocol adaptation layer to reduce the latency of signaling conversion and data transmission and ensure the real-time nature of RIS configuration.

[0295] Error Detection and Correction: Introduce an error detection and correction mechanism to ensure the accuracy and integrity of signaling during protocol conversion and improve the reliability of the system.

[0296] The signaling conversion process in the protocol adaptation layer can be expressed as:

[0297]

[0298] Where: S RIS is the original control signaling of the RIS; is the signaling conversion function; S 5G is the converted 5G NR compatible control signaling. Through the protocol conversion function realize the seamless docking of the RIS control signaling with the existing communication protocol to ensure the collaborative work of the system.

[0299] (3) Modular system architecture

[0300] The modular system architecture adopts a modular design, integrating the RIS as an independent module into the existing base station or network infrastructure, simplifying the system integration process.

[0301] A. Modular design principles:

[0302] Function module division: Divide the RIS system into different function modules, such as signal processing modules, control modules, power management modules, etc., to ensure the functional independence and replaceability of each module.

[0303] Interface standardization: Define standardized interfaces for each function module to ensure interoperability and compatibility between modules and simplify the system integration process.

[0304] B. RIS module integration:

[0305] Base station side integration: Connect the RIS module to the base station through a standardized interface to achieve efficient data exchange and control instruction transmission between the RIS and the base station; Network infrastructure integration: Integrate the RIS module into the existing network infrastructure, such as switches, routers, etc., to form a unified network management and control system.

[0306] C. Modular deployment strategy:

[0307] Distributed deployment: According to communication requirements and coverage, distribute and deploy the RIS modules in different geographical locations to achieve wide signal coverage and enhancement effects; Flexible expansion: According to system requirements, flexibly add or remove RIS modules to adapt to wireless communication environments of different scales and complexities.

[0308] The modular design makes the maintenance and troubleshooting of the RIS system more convenient, reducing system downtime. Through the modular architecture, the RIS system can be easily upgraded in function and improved in performance, extending the service life of the system. The module connection relationship in the modular system architecture can be expressed as a graph structure in graph theory:

[0309] G = (V, E);

[0310] V is a set of modules, including base station modules, RIS modules, user equipment modules, etc.; E is a set of connection relationships between modules, representing the data and control signaling transmission paths between modules. Through the graph structure G, the connection and collaboration relationships between various modules in the modular system architecture are described, realizing the efficient integration and management of the system.

[0311] (4) Standardized optimization algorithm

[0312] The standardized optimization algorithm aims to formulate the standardized processes and interface specifications for the RIS configuration optimization algorithm, ensure the compatibility and replaceability between different optimization algorithms, and promote the wide application and popularization of the algorithms.

[0313] A. Standardization of the optimization algorithm process:

[0314] Input and output specifications: Define the input and output formats of the optimization algorithm, including CSI data format, RIS configuration parameter format, etc., to ensure that different algorithms can run based on a unified data interface; Standardization of algorithm steps: Formulate the standardized steps and processes of the optimization algorithm, such as data preprocessing, feature extraction, strategy optimization, configuration output, etc., to ensure the consistency and repeatability of the algorithm in different systems.

[0315] B. Formulation of interface specifications:

[0316] API design: Design a standardized application programming interface (API) for the RIS configuration optimization algorithm, including function calls, parameter passing, and result return, etc., to ensure the ease of use and interoperability of the algorithm; Plug-in mechanism: Introduce a plug-in mechanism to allow different optimization algorithms to be integrated into the RIS system as plug-in modules, facilitating the switching and expansion of the algorithms.

[0317] C. Compatibility and replaceability:

[0318] Algorithm compatibility testing: Establish an algorithm compatibility testing platform to evaluate the running effects of different optimization algorithms under a unified interface, ensuring the compatibility and interoperability between algorithms; Modular algorithm design: Adopt a modular design concept to design the optimization algorithm as an independent functional module, facilitating the replacement and upgrade of the algorithm.

[0319] D. Standardized documentation and certification:

[0320] Standardized Document Writing: Write a detailed standardized document for the optimization algorithm, covering algorithm processes, interface specifications, data formats, etc., to guide different developers and manufacturers in implementing standardized optimization algorithms; Algorithm Certification Mechanism: Establish an algorithm certification mechanism to certify optimization algorithms that comply with the standardized processes and interface specifications, ensuring their reliability and performance in the RIS system. The interface specification of the standardized optimization algorithm can be expressed in the form of a function:

[0321] Θ = Optimize(CSI, Params);

[0322] Where: Θ is the RIS configuration parameter; CSI is the input channel state information; Params is the parameter configuration of the optimization algorithm. Through the standardized optimization function Optimize(), different optimization algorithms can be interchanged and upgraded based on a unified input and output interface.

[0323] (5) Cross-Layer Cooperative Optimization

[0324] Cross-layer cooperative optimization realizes cross-layer cooperative optimization between RIS and other communication technologies (such as massive multiple-input multiple-output (Massive MIMO), edge computing, etc.), improving the overall system performance and resource utilization efficiency.

[0325] A. Cross-Layer Cooperative Optimization Framework:

[0326] Cooperation between the physical layer and the data link layer: Through RIS, optimize signal beamforming and resource allocation to improve the signal quality of the physical layer and the transmission efficiency of the data link layer; Cooperation between the network layer and the application layer: Use edge computing nodes to process and analyze RIS-optimized data to achieve cooperative optimization between the network layer and the application layer, improving the overall system performance and user experience.

[0327] B. Cooperative Optimization with Massive MIMO:

[0328] Joint beamforming: Combine the beamforming technology of massive MIMO with the reflection beam orientation of RIS to achieve multipath enhancement and interference suppression of signals, improving the system coverage and capacity.

[0329] Resource sharing and scheduling: In a massive MIMO system, RIS acts as an auxiliary device to participate in resource sharing and scheduling, optimizing the utilization efficiency of spectrum and energy resources.

[0330] The joint beamforming model can be expressed as:

[0331] y = H MU-MIMO W MU-MIMO x + H RU ΘH BR x + n;

[0332] where x is the transmitted signal vector, n is the additive noise vector at the receiver, and H MU-MIMO is the multi-user MIMO channel matrix, and W MU-MIMO is the beamforming matrix of multi-user MIMO.

[0333] C. Cooperative Optimization with Edge Computing:

[0334] Data Processing and Analysis: Use edge computing nodes to perform real-time processing and analysis on RIS-optimized data to achieve low-latency RIS configuration optimization decisions; Intelligent Decision Support: Combine the artificial intelligence (AI) and machine learning (ML) capabilities of edge computing to enhance the intelligence and adaptability of the RIS configuration optimization algorithm.

[0335] D. Resource Optimization and Load Balancing:

[0336] Dynamic Resource Allocation: Based on the cross-layer cooperative optimization framework, dynamically adjust the allocation of RIS, MIMO, and edge computing resources to optimize the resource utilization efficiency of the system; Load Balancing Mechanism: Design a load balancing mechanism to ensure resource load balancing among different communication technologies, avoid overload of a certain technology module, and improve the overall stability and performance of the system. The resource allocation model for cross-layer cooperative optimization can be expressed as:

[0337]

[0338] where: W is the beamforming matrix of large-scale MIMO; Θ is the configuration matrix of RIS; SINR k is the signal-to-interference-plus-noise ratio (SINR) of the k-th user, is the total number of users. means to select a pair (Θ, W) from all possible combinations of RIS configuration Θ and beamforming matrix W to maximize the following objective function. By jointly optimizing W and Θ, the signal quality is maximized and the interference is minimized, thus improving the overall performance of the system.

[0339] In summary, the introduction of the unified interface design and protocol adaptation layer ensures high compatibility and interoperability between different manufacturers and devices, promoting the widespread application of RIS technology; the modular system architecture design simplifies the integration process of RIS with base stations and network infrastructure, reducing the complexity and time cost of system deployment, and enhancing the maintainability and scalability of the system; the standardized optimization algorithm process and interface specifications promote the compatibility and replaceability of different optimization algorithms, support the application of diverse optimization strategies in the RIS system, and enhance the intelligence level of the system; cross-layer cooperative optimization achieves efficient cooperation between RIS and other communication technologies, significantly improving the signal quality, capacity, and energy efficiency of the system, meeting the requirements of modern wireless communication for high performance and high reliability; the modular design and cross-layer cooperative optimization enable the RIS system to flexibly adapt to different communication environments and application scenarios, with high adaptability and robustness, ensuring the stable operation of the system in a dynamically changing environment. In summary, through system integration and standardization methods such as unified interface design, protocol adaptation layer, modular system architecture, standardized optimization algorithms, and cross-layer cooperative optimization, the problems of lack of integration and standardization between RIS and existing communication systems have been successfully solved, significantly enhancing the application effect and promotion value of RIS technology in wireless communication systems.

[0340] Please refer to Figure 2 , this embodiment also provides a wireless communication method, which is applied to the sending end and includes: Step S1: Obtain a wireless signal and target channel state information; Step S2: Optimize the target channel state information according to the target optimization model to obtain a RIS configuration instruction, so that the reconfigurable intelligent surface adjusts the reflection units of the reconfigurable intelligent surface based on the RIS configuration instruction, and then the reconfigurable intelligent surface reflects the wireless signal to obtain a reflected signal, and then enables the user equipment to generate an uplink signal according to the reflected signal and send the uplink signal to the sending end.

[0341] The method provided in this embodiment can be executed based on the above-mentioned sending end. The sending end obtains a wireless signal and target channel state information, where the target channel state information provides basic data for the optimal configuration of the reconfigurable intelligent surface; the sending end is used to optimize the target channel state information according to the target optimization model, so as to determine better configurations of each reflection unit in the current reconfigurable intelligent surface, generate a more reliable current RIS configuration instruction, so that the reconfigurable intelligent surface adjusts the reflection units based on the RIS configuration instruction, so that when the reconfigurable intelligent surface reflects the wireless signal transmitted by the sending end, better signal enhancement and interference suppression can be achieved, and a more reliable reflected signal can be obtained to be sent to the user equipment. And the sending end generates a RIS configuration instruction that better conforms to the current state in real time based on the target channel state information, meeting the requirements of the dynamically changing wireless environment and improving the reliability of wireless communication.

[0342] The above are the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made, and these improvements and modifications are also regarded as the protection scope of the present invention.

Claims

1. A wireless communication system, characterized in that, Including: A transmitting end, configured to obtain a wireless signal and target channel state information, and optimize the target channel state information according to a target optimization model to obtain a RIS configuration instruction, so that a reconfigurable intelligent surface adjusts reflection units of the reconfigurable intelligent surface based on the RIS configuration instruction, and further enables the reconfigurable intelligent surface to reflect the wireless signal to obtain a reflected signal; A user equipment, configured to generate an uplink signal according to the reflected signal and send the uplink signal to the transmitting end.

2. The wireless communication system according to claim 1, characterized in that, The transmitting end is configured to obtain a wireless signal and target channel state information, including: When obtaining the target channel state information, receiving environment perception data and direct channel state information sent by the user equipment; Based on the direct channel state information and the environment perception data, constructing RIS channel information through a preset channel inference algorithm; Based on the direct channel state information and the RIS channel information, constructing the target channel state information.

3. The wireless communication system according to claim 1, wherein The transmitting end is configured to obtain a wireless signal and target channel state information, including: When obtaining the target channel state information, compressing the target channel state information based on a compressive sensing algorithm to obtain low-dimensional information; Reconstructing the low-dimensional information based on a signal reconstruction algorithm to obtain the target channel state information.

4. The wireless communication system according to claim 3, characterized in that, When the transmitting end is configured to obtain the target channel state information, compressing the target channel state information based on a compressive sensing model to obtain low-dimensional information, including: Obtaining a random measurement matrix, where the number of rows of the random measurement matrix is less than the number of rows of the channel matrix of the target channel state information; Performing linear projection on the target channel state information through the random measurement matrix to obtain low-dimensional information.

5. The wireless communication system according to claim 1, characterized in that, The transmitting end optimizes the target channel state information according to the target optimization model to obtain a RIS configuration instruction, including: Obtaining sample configuration data, where sample state information in the sample configuration data is used as input information, and sample action information in the sample configuration data is used as output information; Based on the sample configuration data, training a preset optimization model through a reward function to obtain the target optimization model; Obtaining to-be-measured state information, where the to-be-measured state information includes the target channel state information; Processing the to-be-measured state information through the target optimization model to obtain target action information; Generating the RIS configuration instruction based on the target action information.

6. The wireless communication system according to claim 5, characterized in that, The transmitting end is configured to train a preset optimization model through a reward function based on the sample configuration data to obtain the target optimization model, including: Based on a signal-to-noise ratio algorithm and the sample state information, obtaining the signal-to-noise ratio of the user equipment; Based on the received power of interference sources at different times, obtaining the signal-to-noise ratio of the interference sources; Based on the signal-to-noise ratio of the user equipment and the signal-to-noise ratio of the interference sources, constructing the reward function.

7. The wireless communication system according to claim 1, characterized in that, The transmitting end is configured to optimize the target channel state information according to the target optimization model, including: Based on a target prediction and completion model, predicting and completing the target channel state information to obtain complete channel state information; Optimize the complete channel state information according to the target optimization model to obtain the RIS configuration instruction.

8. The wireless communication system according to claim 7, wherein, The transmitter is used to predict and complete the target channel state information based on the target prediction completion model to obtain the complete channel state information, including: Obtain the sample channel state information; Train the preset prediction completion model through the sample channel state information to obtain the target prediction completion model; Input the target channel state information into the target prediction completion model to obtain the complete channel state information.

9. The wireless communication system according to claim 8, characterized in that, The transmitter is used to predict and complete the target channel state information based on the target prediction completion model to obtain the complete channel state information, including: Divide the target prediction completion model into a training set and a test set; Train the preset prediction completion model through the training set to obtain the initial trained prediction completion model; Input the test set into the initial trained prediction completion model to obtain the predicted value; Evaluate the error of the predicted value to obtain the prediction error. When the prediction error is greater than the preset threshold, adjust the parameters of the initial trained prediction completion model until the prediction error is less than or equal to the preset threshold, and output the target prediction completion model.

10. A wireless communication method, characterized in that, Applied to the transmitter, including: Obtain the wireless signal and the target channel state information; Optimize the target channel state information according to the target optimization model to obtain the RIS configuration instruction, so that the reconfigurable intelligent surface adjusts the reflection unit of the reconfigurable intelligent surface based on the RIS configuration instruction, and then the reconfigurable intelligent surface reflects the wireless signal to obtain the reflected signal; and then enable the user equipment to generate an uplink signal according to the reflected signal and send the uplink signal to the transmitter.

Images