A communication method, device and system based on reconfigurable intelligent surface

Abstract

This invention discloses a communication method, apparatus, and system based on reconfigurable smart surfaces, belonging to the field of wireless communication technology. The communication method based on reconfigurable smart surfaces jointly optimizes the transmitter's current beamforming vector and current RIS reflection parameters based on real-time channel estimation results using a deep deterministic policy gradient algorithm. A joint optimization model of the transmitter's beamforming and RIS parameters is established, transforming the non-convex optimization problem into a Markov decision process, which is efficiently solved using the deep deterministic policy gradient algorithm, thereby improving the transmission capability of the communication. Furthermore, the introduction of reconfigurable smart surface technology, with its design featuring a reflection unit possessing independent amplitude-phase modulation capability, overcomes the fixed amplitude limitation of traditional passive RIS, effectively overcoming dual-fading path loss, and thus improving the anti-interference performance of the communication system in complex electromagnetic environments such as high spectral density and high interference power.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, and more specifically, relates to a communication method, apparatus and system based on a reconfigurable smart surface. Background Technology

[0002] With the rapid development of mobile communication technology, challenges related to communication security and privacy are gradually becoming one of the key issues that 6G (6th Generation Mobile Networks) systems aim to address. Traditional solutions proposed by the industry to address this issue include frequency hopping, spread spectrum technology, and multi-antenna technology.

[0003] Frequency hopping technology rapidly changes the carrier frequency of the transmitted signal, making it difficult for interferers to predict and track, thus achieving anti-interference. However, this method places high demands on the management and allocation of frequency resources and has limited effectiveness in high-density spectrum environments. Spread spectrum technology spreads the signal across a wider frequency band, dispersing the energy of the interfering signal over a wider area, thereby reducing the impact of interference. Although spread spectrum technology performs well in anti-interference, its transmission rate is relatively low and it may still fail under high-power interference. Multi-antenna technology utilizes multiple antennas for signal transmission and reception, improving the system's anti-interference capability through spatial diversity and beamforming techniques. However, multi-antenna systems have high hardware and computing resource requirements and are susceptible to multipath effects in complex electromagnetic environments.

[0004] In other words, existing communication systems have poor anti-interference performance in complex electromagnetic environments such as high spectral density and high interference power. Summary of the Invention

[0005] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a communication method, device and system based on reconfigurable smart surface, the purpose of which is to solve the technical problem of poor anti-interference performance of existing communication systems in complex electromagnetic environments such as high spectral density and high interference power.

[0006] To achieve the above objectives, according to one aspect of the present invention, a communication method based on a reconfigurable smart surface is provided, comprising:

[0007] S1: Set the initial reflection coefficient matrix of the reconfigurable smart surface RIS and the initial beamforming vector of the transmitter;

[0008] S2: Control the transmitter to transmit a first communication signal via the first direct link and the first RIS auxiliary link corresponding to the initial reflection coefficient matrix using the initial beamforming vector; control the interference source to transmit a first interference signal via the second direct link and the second RIS auxiliary link corresponding to the initial reflection coefficient matrix; control the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the first communication signal and the first interference signal in real time;

[0009] S3: The current beamforming vector and the current RIS reflection parameters of the transmitter are jointly optimized based on the real-time channel estimation results using a deep deterministic policy gradient algorithm. The deep deterministic policy gradient algorithm includes: a reward function with the maximum signal-to-interference-plus-noise ratio of the received signal as the core; a Markov decision process state space containing channel state information, power constraints, and historical performance indicators; and a continuous action space containing the beamforming vector, RIS phase offset, and amplitude gain.

[0010] S4: Control the transmitter to transmit a second communication signal via the initial beamforming vector through the first direct link and the third RIS auxiliary link corresponding to the current reflection coefficient matrix; control the interference source to transmit a second interference signal via the second direct link and the fourth RIS auxiliary link corresponding to the current reflection coefficient matrix; control the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the second communication signal and the second interference signal in real time;

[0011] S5: If the adjustment period has not been reached, proceed directly to S3; if the adjustment period has been reached, dynamically update the reflection coefficient matrix obtained in S4 according to the channel change rate, and then proceed to S3.

[0012] Furthermore, the reward function r t Designed as: r t =10log 10 (γ t )-λ1Δp t -λ2D t ; where γ t Let ΔP be the signal-to-interference-plus-noise ratio (SIR) of the received signal at time t. t Let D be the power constraint violation at time t. t Let λ be the service quality deviation at time t, and λ1 and λ2 be two penalty coefficients.

[0013] Furthermore, the state space s t Indicates: s t =[vec(G l ),h j ,h,P t ,P j ,γ t-1 The action space a tIndicates: a t =[w t ,θ1,…,θ N ,β1,…,β N ]; where vec(G l ) is the vector corresponding to the channel matrix of the sender-RIS, h j Let h be the channel matrix from the interference source to the RIS, and h be the channel matrix from the RIS to the receiver. t P represents the transmitter's transmit power. j γ is the transmission power of the interference source. t-1 Let w be the signal-to-interference-plus-noise ratio (SIR) of the received signal at time t-1. t Let θ1,…,θ be the square beamforming vector transmitted at time t. N Let β1,…,β be the phase shift of each unit in the RIS at time t. N Let be the amplitude change of the RIS units at time t.

[0014] Furthermore, the optimization process of the current RIS reflection parameters in the deep deterministic strategy gradient algorithm is as follows: first optimize the amplitude gain parameters based on gradient descent, and then perform amplitude-phase joint optimization under power constraints.

[0015] Furthermore, the amplitude-phase joint optimization satisfies the following constraint: Where N is the total number of reflective units, β n ∈[1,β max [] represents the amplitude gain coefficient of the nth reflecting unit. Let P be the incident power of the nth reflecting element. c For circuit power consumption, This represents the maximum available power for the RIS.

[0016] Furthermore, the amplitude-phase joint optimization under the reimplementation of power constraints includes: using the current optimal phase A dynamic search window is established around the center; the window width Δθ is adjusted according to the direction of signal-to-interference-plus-noise ratio change in adjacent subframes to obtain the interval. A local fine-grained search is triggered when the window width is less than π / 12.

[0017] Furthermore, the step of dynamically updating the reflection coefficient matrix obtained in S4 according to the channel change rate when the adjustment period is reached includes: pre-correcting the phase deviation caused by Doppler frequency shift; dynamically adjusting the gain coefficient in the reflection coefficient matrix according to the received signal strength; and actively probing the channel state information with the transmit beam for non-cooperative interference.

[0018] According to another aspect of the present invention, a communication device based on a reconfigurable smart surface is provided, comprising:

[0019] The initial setup module is used to set the initial reflection coefficient matrix of the reconfigurable smart surface RIS and the initial beamforming vector of the transmitter.

[0020] The first control module is used to control the transmitter to transmit a first communication signal via the initial beamforming vector through the first direct link and the first RIS auxiliary link corresponding to the initial reflection coefficient matrix; the interference source to transmit a first interference signal via the second direct link and the second RIS auxiliary link corresponding to the initial reflection coefficient matrix; and the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the first communication signal and the first interference signal in real time.

[0021] The iterative optimization module is used to jointly optimize the current beamforming vector and the current RIS reflection parameters of the transmitter based on the channel estimation results acquired in real time using a deep deterministic policy gradient algorithm. The deep deterministic policy gradient algorithm includes: a reward function with the core of maximizing the signal-to-interference-plus-noise ratio of the received signal; a Markov decision process state space containing channel state information, power constraints, and historical performance indicators; and a continuous action space containing the beamforming vector, RIS phase offset, and amplitude gain.

[0022] The second control module is used to control the transmitter to transmit a second communication signal via the first direct link and the third RIS auxiliary link corresponding to the current reflection coefficient matrix using the initial beamforming vector; the interference source to transmit a second interference signal via the second direct link and the fourth RIS auxiliary link corresponding to the current reflection coefficient matrix; and the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the second communication signal and the second interference signal in real time.

[0023] The iterative judgment module is used to directly switch to the loop optimization module when the adjustment period has not been reached; when the adjustment period is reached, the reflection coefficient matrix output by the second control module is dynamically updated according to the channel change rate, and then the loop optimization module is switched to.

[0024] According to another aspect of the present invention, a communication system based on a reconfigurable smart surface is provided, comprising: a transmitter, a reconfigurable smart surface RIS, an interference source, a receiver, a memory, and a processor. The transmitter transmits a communication signal to the receiver via its corresponding RIS auxiliary link and a direct link, while the interference source transmits an interference signal to the receiver via its corresponding RIS auxiliary link and a direct link. The receiver measures the signal-to-interference-plus-noise ratio (SIR) of the received signals corresponding to the communication signal and the interference signal in real time. The memory stores a computer program, and the processor executes the computer program to implement the steps of the communication method based on the reconfigurable smart surface.

[0025] According to another aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the communication method based on the reconfigurable smart surface.

[0026] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0027] This invention provides a communication method based on reconfigurable smart surfaces. It utilizes a deep deterministic policy gradient algorithm to jointly optimize the sender's current beamforming vector and current RIS reflection parameters based on real-time channel estimation results. A joint optimization model of the sender's beamforming and RIS parameters is established, transforming the non-convex optimization problem into a Markov decision process, which is efficiently solved using the deep deterministic policy gradient algorithm, thereby improving the communication transmission capability. Furthermore, reconfigurable smart surface technology is introduced, featuring reflector units with independent amplitude-phase adjustment capabilities, overcoming the fixed amplitude limitation of traditional passive RIS. The gain of each reflector unit is adjustable from 1 to 3 times, effectively overcoming dual-fading path loss, thus improving the anti-interference performance of the communication system in complex electromagnetic environments with high spectral density and high interference power. Attached Figure Description

[0028] Figure 1 This is a flowchart of a communication method based on a reconfigurable smart surface provided in Embodiment 1 of the present invention;

[0029] Figure 2 A schematic diagram showing the result of the reconfigurable smart surface-assisted communication system provided in Embodiment 1 of the present invention;

[0030] Figure 3 This is a flowchart of the heterogeneous resource collaborative scheduling method based on fog node cooperation provided in Embodiment 1 of the present invention;

[0031] Figure 4 A flowchart illustrating the formulation of an ADMM-based caching strategy provided in Embodiment 1 of the present invention;

[0032] Figure 5 This is a schematic diagram of a heterogeneous resource collaborative scheduling system based on fog node cooperation provided in Embodiment 1 of the present invention;

[0033] Figure 6 This is a schematic diagram illustrating how a heterogeneous resource collaborative scheduling system based on fog node cooperation provides services to users, as shown in Embodiment 1 of the present invention.

[0034] Figure 7 This is a simulation diagram of the convergence performance of the ADMM-based caching algorithm provided in Embodiment 1 of the present invention.

[0035] Figure 8The graph shows the relationship between the optimal average latency and heterogeneous network resources provided in Embodiment 1 of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0037] Example 1

[0038] This embodiment provides a communication method based on a reconfigurable smart surface, such as... Figure 1 As shown, it includes: S1-S5. The communication method is applied to RIS-assisted communication systems, such as... Figure 2 As shown, it includes a transmitter, a Reflection Array (RIS), an interference source, a receiver, a memory, and a processor. The RIS can be active or passive. Taking an active RIS as an example, an active RIS contains N independently adjustable reflection units, each with dual phase and amplitude adjustment capabilities. The active RIS reflection model is designed with reflection units possessing independent amplitude-phase adjustment capabilities, breaking through the fixed amplitude limitation of traditional passive RISs. The gain of each reflection unit is adjustable from 1 to 3 times, effectively overcoming dual-path fading loss. Furthermore, the active RIS, an integrated reconfigurable smart surface with active amplifier circuitry, overcomes the dual-path fading defect of traditional passive RISs by independently adjusting the phase (θ∈[0,2π)) and amplitude (β∈[1,3]) of the reflected signal, possessing signal amplification capabilities and dynamically adapting to high-interference environments. The transmitter sends communication signals to the receiver via its corresponding RIS auxiliary link and direct link, while the interference source sends interference signals to the receiver via its corresponding RIS auxiliary link and direct link; the receiver measures the signal-to-interference-plus-noise ratio (SNR) of the received signals corresponding to the communication signal and the interference signal in real time. Its advantages lie in its intelligent anti-interference mechanism: it constructs a closed-loop system that includes interference feature extraction, reflected beamforming, and signal cancellation, enabling real-time perception and dynamic cancellation of interference signals.

[0039] S1: Set the initial reflection coefficient matrix of the reconfigurable smart surface (RIS) and the initial beamforming vector of the transmitter. Specifically, initialize the reflection coefficient matrix of the active RIS, set the initial phase of each reflection element to be uniformly distributed in the interval [0, 2π], and set the initial amplitude value to the preset maximum gain. Further, the reflection coefficient matrix Φ of the active RIS is expressed as: Where β n ∈[1,β max ] represents the amplitude gain coefficient of the nth reflecting unit, θ n∈[0,2π) represents the phase offset, β max This is the preset maximum gain value.

[0040] S2: The transmitter uses the initial beamforming vector to transmit the first communication signal via the first direct link and the first RIS auxiliary link corresponding to the initial reflection coefficient matrix; the interference source transmits the first interference signal via the second direct link and the second RIS auxiliary link corresponding to the initial reflection coefficient matrix; the receiver uses the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the first communication signal and the first interference signal in real time.

[0041] S3: The deep deterministic policy gradient algorithm jointly optimizes the current beamforming vector and the current RIS reflection parameters of the transmitter based on the channel estimation results obtained in real time. The deep deterministic policy gradient algorithm includes: a reward function with the core of maximizing the signal-to-interference-plus-noise ratio of the received signal; a Markov decision process state space containing channel state information, power constraints and historical performance indicators; and a continuous action space containing beamforming vector, RIS phase offset and amplitude gain.

[0042] Among them, the Deep Deterministic Policy Gradient (DPRQ) algorithm, a reinforcement learning algorithm for continuous action spaces, consists of an Actor-Critic dual network. The Actor network outputs a deterministic policy (such as RIS phase / amplitude parameters), while the Critic network evaluates the Q-value. Real-time optimization in dynamic channel environments is achieved through empirical replay and soft update mechanisms, resulting in a significantly faster convergence speed compared to traditional algorithms.

[0043] Specifically, the deep deterministic policy gradient algorithm employs a dual-network architecture, including a policy network: an input state vector with dimension (M+3N+2) and an output action vector with dimension (M+2N); and an evaluation network: an input joint state-action vector with dimension (M+3N+2+M+2N) and an output Q-value evaluation; where M is the number of transmitting antennas and N is the number of RIS reflector units. Furthermore, parameter adaptation is achieved through a target network soft update mechanism, including: setting an experience replay buffer to store historical state transition data, using Ornstein-Uhlenbeck noise to enhance exploration capabilities, and performing periodic parameter synchronization to ensure algorithm stability.

[0044] S4: The transmitter uses the initial beamforming vector to transmit the second communication signal via the first direct link and the third RIS auxiliary link corresponding to the current reflection coefficient matrix; the interference source transmits the second interference signal via the second direct link and the fourth RIS auxiliary link corresponding to the current reflection coefficient matrix; the receiver uses the receiver to measure the signal-to-interference-plus-noise ratio of the received signal corresponding to the second communication signal and the second interference signal in real time.

[0045] S5: If the adjustment period has not been reached, proceed directly to S3; if the adjustment period has been reached, dynamically update the reflection coefficient matrix obtained in S4 according to the channel change rate, and then proceed to S3. Specifically, the reflection coefficient matrix output by the joint optimization module can be dynamically updated according to the channel change rate ρ, and the adjustment period T satisfies:

[0046] Furthermore, the following algorithm can be used to determine the key performance indicators of the communication method, including: anti-interference gain: Δγ = γ RIS -γ direct Power efficiency: Convergence time: The time required to reach 95% of the maximum SINR.

[0047] Furthermore, the reward function r t Designed as: r t =10log 10 (γ t )-λ1ΔP t -λ2D t ;γ t Let ΔP be the signal-to-interference-plus-noise ratio (SIR) of the received signal at time t. t Let D be the power constraint violation at time t. t Let be the service quality deviation at time t, and λ1 and λ2 be two penalty coefficients. Further, the state space s t Indicates: s t =[vec(G l ),h j ,h,P t ,P j ,γ t-1 Action space a t Indicates: a t =[w t ,θ1,…,θ N ,β1,…,β N ]. Among them, vec(G l ) is the vector corresponding to the channel matrix of the sender-RIS, h j Let h be the channel matrix from the interference source to the RIS, and h be the channel matrix from the RIS to the receiver. t P represents the transmitter's transmit power. j γ is the transmission power of the interference source. t-1 Let w be the signal-to-interference-plus-noise ratio (SIR) of the received signal at time t-1. t Let θ1,…,θ be the square beamforming vector transmitted at time t. N Let β1,…,β be the phase shift of each unit in the RIS at time t. N Let be the amplitude change of the RIS units at time t.

[0048] Furthermore, the optimization process for the current RIS reflection parameters in the deep deterministic policy gradient algorithm is as follows: first, the amplitude gain parameter is optimized based on gradient descent, and then amplitude-phase joint optimization is performed under power constraints. Furthermore, the amplitude-phase joint optimization satisfies the following constraint: N is the total number of reflective units, β n ∈[1,β max [] represents the amplitude gain coefficient of the nth reflecting unit. Let P be the incident power of the first n-th reflecting unit. c For circuit power consumption, To determine the maximum available power for the RIS, further, the amplitude-phase joint optimization under power constraints includes: employing an improved golden section method to achieve the current optimal phase. A dynamic search window is established around the center; the window width Δθ is adjusted according to the direction of signal-to-interference-plus-noise ratio change in adjacent subframes to obtain the interval. A local fine-grained search is triggered when the window width is less than π / 12. Its advantages lie in designing a parameter update strategy based on the channel change rate, employing an improved golden section method for fast phase search, and combining gradient descent to optimize the amplitude parameter.

[0049] Furthermore, when the adjustment period is reached, the reflection coefficient matrix obtained in S4 is dynamically updated according to the channel change rate, including: pre-correction for phase deviation caused by Doppler frequency shift; dynamic adjustment of the gain coefficient in the reflection coefficient matrix according to the received signal strength; and active detection of channel state information by the transmit beam for non-cooperative interference.

[0050] The following describes a process of heterogeneous resource collaborative scheduling based on fog node cooperation using the communication method provided in this embodiment:

[0051] The first step considers a RIS-assisted Multiple-Input Single-Output (MISO) communication anti-jamming scenario. A transmitter equipped with M antennas communicates downlink with a receiver equipped with a single antenna, while a malicious jamming source equipped with L antennas interferes with the receiver across the entire frequency band. The RIS consists of N passive reflector elements, controlled by the transmitter through a RIS controller. This paper comprehensively considers a fully connected link propagation scenario where both direct and reflected links exist simultaneously; that is, legitimate signals and jamming signals reach the receiver through the direct link and the RIS-assisted reflected link, respectively.

[0052] Since path loss significantly impacts signal strength, it is assumed that the signal power reaching the RIS after multiple reflections is negligible. To improve communication anti-interference performance, it is assumed that Channel State Information (CSI) at each communication node can be accurately detected and acquired. For the CSI of legitimate signal propagation channels, it is obtained through conventional pilot estimation methods; while for the CSI of non-cooperative interference signal propagation channels, it is obtained by deploying a small number of active detection sensors at the RIS and receiver. The effective baseband channel models from transmitter to RIS, from interference source to RIS, from transmitter to receiver, from interference source to receiver, and from RIS to receiver are respectively expressed as: G l =g L {h nm} n∈[1,N],m∈[1,M] ∈C N×M G j =g J {h nl} n∈[1,N],l∈[1,L] ∈C L×N ,

[0053] Without loss of generality, we assume that all channels H follow a Rician channel model, i.e. k is the Rice factor (K-factor), which characterizes the power ratio of the LoS path to the scattering path, H LOS For deterministic line-of-sight path components; H NLOS The scattering components are Rayleigh distributed and follow a complex Gaussian distribution. Ignoring RIS units used for other functions, and assuming all RIS reflective units are used for anti-interference, the reflection coefficient matrix of the RIS can be expressed as Φ. l =(Γ l,1 ,Γ l,2 ,……,Γ l,N )∈C N×N Φ j =(Γ j,1 ,Γ j,2 ,……,Γ j,N )∈C N×N ,in and Let α and β represent the reflection coefficients of the nth reflecting unit for the legitimate signal and the interference signal, respectively, with amplitude variations of α and β. l,n and α j,n The phase shifts are respectively and These are the transmit beamforming vectors of the transmitter and the interfering source, respectively. By weighting and adjusting the antenna phase and amplitude, the beamforming vectors of the transmitter and the interfering source must satisfy the power constraint w. l <P l and w j <P l Through power P l and P j Achieve standardization, that is

[0054] Based on the actual signal transmission process, the received signal can be represented as:

[0055]

[0056] This indicates a legitimate signal from the receiver. Let n represent the interference signal at the receiver, and n be the variance of the receiver's white Gaussian noise. Therefore, the SINR of the receiver's legitimate signal can be expressed as:

[0057]

[0058] To improve the anti-interference capability of the considered communication system, under multiple constraints such as transmit power and RIS reflection phase shift, a maximum legitimate signal SINR optimization problem is constructed for the transmit beamforming vector and RIS reflection coefficient matrix, as follows:

[0059]

[0060] stC1:||w l || 2 ≤P l,max

[0061]

[0062] C3:S U ≥S min

[0063] Among them, constraint C1 is a transmit power constraint on the transmitter's beamforming vector; specifically, the transmitter's transmit power must not exceed P. l,max C2 is the phase shift constraint on the RIS reflection coefficient matrix, assuming that the signal amplitude remains unchanged after reflection, and adjusting the continuous phase shift to be between (0, 2π]. C3 is the receiver sensitivity requirement, ensuring that the receiver's signal quality meets the relevant Quality of Service (QoS) requirements.

[0064] Since the interference source emits full-band interference, while the legitimate signal is only on a certain frequency band, it can be assumed that the interference and legitimate signal are always in the same frequency band, and the RIS has the same reflection matrix for both. Therefore, assume Φ l =Φ j Clearly, from the optimization problem, we can see that the optimization variable w l and Φ l The dimensions are high and the two are tightly coupled. The constraint on the RIS phase shift is a non-convex rank-1 constraint, and the objective function is a non-convex fractional function.

[0065] The second step involves solving the aforementioned optimization problem to obtain the optimal signal-to-interference-plus-noise ratio (SINR). This problem is a fractional nonlinear problem, typically NP-hard. A high degree of nonlinear coupling exists between the beamforming vector and the RIS reflection matrix, potentially making the problem inherently nonconvex. Channel parameters exhibit slow time-varying behavior (at a rate of α), requiring a real-time adaptive strategy. However, each channel change necessitates solving an optimization problem, leading to computational delays that violate real-time operational constraints. Considering the nonconvexity of the problem and the dynamic changes in the channel, traditional convex optimization methods exhibit significant limitations in solving this problem. DDPG, with its ability to handle continuous actions, dynamic adaptability, and nonconvex optimization, significantly outperforms convex optimization in active RIS anti-jamming systems. Especially in high-dimensional RIS configurations, rapidly changing channels, and hardware-constrained scenarios, DDPG can achieve optimizations with lower latency and higher anti-jamming gain.

[0066] The MDP framework for the interference resistance problem includes four tuples: state space, action space, state transition probabilities, and reward function.

[0067] 1) State space s: The state space needs to comprehensively represent dynamic parameters, including: CSI; RIS configuration state; characteristics and location of the jammer; system resources, including system power supply, RIS energy efficiency, etc.

[0068] 2) Action space a: The action needs to implement RIS reflection parameter adjustment Φ l and BS beamforming vector w l Joint optimization.

[0069] 3) State transition probability p: State transitions require modeling the randomness of the communication environment and RIS actions. Based on the Ricean fading model, the time-varying characteristics of the channel gain are considered.

[0070] 4) Reward Function r: The reward function needs to balance communication performance and resource consumption to maximize anti-interference performance. A dynamic penalty mechanism exists to apply a large negative reward to cases exceeding the constraints of the optimization problem.

[0071] like Figure 3As shown, DDPG combines deep neural networks with deterministic policy gradients, specifically designed for continuous action space optimization, making it suitable for high-dimensional parameter dynamic adjustment problems in anti-jamming communication. In the proposed system, the action space involves continuous parameters, such as phase / amplitude adjustment of RIS reflector units, beamforming vector optimization, and power control. Traditional discretization methods may lead to the curse of dimensionality, while DDPG directly outputs deterministic actions through the participant network. Furthermore, DDPG combines a replay buffer and a target network, effectively addressing non-stationarity issues such as time-varying channel characteristics in dynamic interference environments, and improving sample utilization through offline policy learning.

[0072] The invention will be described in detail below with reference to simulation: The simulation parameters are set as follows: The large-scale fading model of the channel is as follows: Let γ0-30dB represent the path loss factor at a reference distance d0 = 1m, where d represents the distance and α represents the path loss exponent. Without loss of generality, we assume α BR =2.2, α RU =αRE, k=2.5, and α BU =α BE =3.6. G l and G j It follows Rayleigh fading. Unless otherwise specified, we assume M = 8, N = 100, L = 8, and P... l =50dBm, P j =50dBm, P Rm =40dBm ρ = 0.1.

[0073] like Figure 4 As shown, the learning rate plays a crucial role in the DDPG optimization of an active RIS system. A learning rate of 0.001 achieves a near-optimal balance between convergence speed, stability, and robustness. However, an excessively high learning rate may lead to an increase in the gradient norm of the Critic network's loss function, causing oscillations.

[0074] Figure 5 This paper describes the nonlinear relationship between the channel rate of change ρ and the signal-to-interference-plus-noise ratio (SINR), revealing a key system design trade-off: RIS can significantly improve SINR in quasi-static channels (ρ < 0.1); however, its effectiveness diminishes in rapidly changing scenarios. This highlights the need for adaptive algorithms and hardware acceleration to maintain optimization agility in practical time-sensitive application deployments. This nonlinear characteristic essentially stems from the competition between channel coherence time and RIS optimization delay.

[0075] Figure 6 , Figure 7 , Figure 8The results show that the proposed DDPG optimization scheme is close to the genetic algorithm scheme in terms of anti-interference capability and outperforms other optimization schemes. The active RIS enhances the auxiliary capability of the RIS auxiliary link compared to the passive RIS. Furthermore, since the RIS is located closer to the interference source, its interference suppression effect plays a major role. Therefore, minimizing interference is a more effective solution than maximizing the legitimate signal. Figure 6 The results show that as the number of RIS units N increases, the system's anti-interference capability gradually improves, thanks to its better utilization of multipath propagation. Active RIS units have limited growth potential due to power constraints; passive RIS units theoretically have greater growth potential, but are practically limited by hardware costs and other factors. Figure 7 This indicates that the anti-interference capability of an active RIS-assisted system exhibits an initial enhancement followed by saturation as the number of transmit antennas M increases. Its ultimate diversity gain limit is determined by channel independence, hardware limitations, and interference dynamics. When the number of transmit antennas exceeds the channel degrees of freedom or the RIS control dimension, the spatial correlation of multipath propagation leads to a decrease in the marginal benefit of diversity gain. Figure 8 The SINR is shown to vary with the legal transmit power P. l Linear growth, with its directly increasing legal power, enhances the transmission energy of the useful signal. RIS, through beamforming and interference suppression, more effectively converts this energy into signal gain at the receiver. From a practical perspective, the sustainability of this linear gain is limited by the accuracy of RIS hardware reciprocity calibration and the saturation threshold of nonlinear devices.

[0076] Based on the experimental results above, DDPG demonstrates unique advantages in optimizing active RIS-assisted anti-interference systems, especially under dynamically changing channels. DDPG avoids the "curse of dimensionality" caused by combinatorial explosion in high-dimensional action spaces, thus reducing algorithm complexity. It directly outputs continuous RIS parameters through its deterministic policy network, achieving gradient-based optimization of non-convex SINR surfaces.

[0077] Example 2

[0078] This embodiment provides a communication device based on a reconfigurable smart surface, comprising:

[0079] The initial setup module is used to set the initial reflection coefficient matrix of the reconfigurable smart surface RIS and the initial beamforming vector of the transmitter.

[0080] The first control module is used to control the transmitter to transmit a first communication signal via the first direct link and the first RIS auxiliary link corresponding to the initial reflection coefficient matrix using the initial beamforming vector; the interference source to transmit a first interference signal via the second direct link and the second RIS auxiliary link corresponding to the initial reflection coefficient matrix; and the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the first communication signal and the first interference signal in real time.

[0081] The loop optimization module is used to jointly optimize the current beamforming vector and the current RIS reflection parameters of the transmitter based on the channel estimation results obtained in real time through a deep deterministic policy gradient algorithm. The deep deterministic policy gradient algorithm includes: a reward function with the core of maximizing the signal-to-interference-plus-noise ratio of the received signal; a Markov decision process state space containing channel state information, power constraints and historical performance indicators; and a continuous action space containing beamforming vector, RIS phase offset and amplitude gain.

[0082] The second control module is used to control the transmitter to transmit the second communication signal via the first direct link and the third RIS auxiliary link corresponding to the current reflection coefficient matrix using the initial beamforming vector; the interference source to transmit the second interference signal via the second direct link and the fourth RIS auxiliary link corresponding to the current reflection coefficient matrix; and to use the receiver to measure the signal-to-interference-plus-noise ratio of the received signal corresponding to the second communication signal and the second interference signal in real time.

[0083] The iterative judgment module is used to directly switch to the loop optimization module when the adjustment period has not been reached; when the adjustment period is reached, the reflection coefficient matrix output by the second control module is dynamically updated according to the channel change rate, and then the loop optimization module is switched to.

[0084] Example 3

[0085] This embodiment provides a communication system based on a reconfigurable smart surface, including: a transmitter, a reconfigurable smart surface RIS, an interference source, a receiver, a memory, and a processor. The transmitter sends communication signals to the receiver via its corresponding RIS auxiliary link and direct link, while the interference source sends interference signals to the receiver via its corresponding RIS auxiliary link and direct link. The receiver measures the signal-to-interference-plus-noise ratio (SIR) of the received signals corresponding to the communication signal and the interference signal in real time. The memory stores a computer program, and the processor executes the computer program to implement the steps of the communication method based on the reconfigurable smart surface.

[0086] Example 4

[0087] This embodiment provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of a communication method based on a reconfigurable smart surface.

[0088] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A communication method based on reconfigurable intelligent surface, characterized in that, include: S1: Set the initial reflection coefficient matrix of the reconfigurable smart surface RIS and the initial beamforming vector of the transmitter; S2: Control the transmitter to transmit a first communication signal via the first direct link and the first RIS auxiliary link corresponding to the initial reflection coefficient matrix using the initial beamforming vector; control the interference source to transmit a first interference signal via the second direct link and the second RIS auxiliary link corresponding to the initial reflection coefficient matrix; control the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the first communication signal and the first interference signal in real time; S3: The current beamforming vector and the current RIS reflection parameters of the transmitter are jointly optimized based on the real-time channel estimation results using a deep deterministic policy gradient algorithm. The deep deterministic policy gradient algorithm includes: a reward function with the maximum signal-to-interference-plus-noise ratio of the received signal as the core; a Markov decision process state space containing channel state information, power constraints, and historical performance indicators; and a continuous action space containing the beamforming vector, RIS phase offset, and amplitude gain. S4: Control the transmitter to transmit a second communication signal via the initial beamforming vector through the first direct link and the third RIS auxiliary link corresponding to the current reflection coefficient matrix; control the interference source to transmit a second interference signal via the second direct link and the fourth RIS auxiliary link corresponding to the current reflection coefficient matrix; control the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the second communication signal and the second interference signal in real time; S5: If the adjustment period has not been reached, proceed directly to S3; if the adjustment period has been reached, dynamically update the reflection coefficient matrix obtained in S4 according to the channel change rate, and then proceed to S3. The optimization process of the current RIS reflection parameters in the deep deterministic strategy gradient algorithm is as follows: first optimize the amplitude gain parameters based on gradient descent, and then perform amplitude-phase joint optimization under power constraints. The amplitude-phase joint optimization satisfies the following constraint: ; Where N is the total number of reflective units, ∈[1, [] represents the amplitude gain coefficient of the nth reflecting unit. For the first The incident power of each reflecting element For circuit power consumption, This represents the maximum available power of the RIS. The amplitude-phase joint optimization under the reimplementation of power constraints includes: using the current optimal phase A dynamic search window is established around the center; the window width is adjusted according to the direction of signal-to-interference-plus-noise ratio (SIR) change in adjacent subframes. , obtain the interval A local fine-grained search is triggered when the window width is less than π / 12.

2. The communication method based on reconfigurable smart surfaces as described in claim 1, characterized in that, The reward function Designed as follows: ;in, Let be the signal-to-interference-plus-noise ratio (SIR) of the received signal at time t. Let be the power constraint violation at time t. Let be the service quality deviation at time t. , These are two penalty coefficients.

3. The communication method based on reconfigurable smart surfaces as described in claim 2, characterized in that, The state space express: ; The action space express: ; in, This is the vector corresponding to the channel matrix of the sender-RIS. The channel matrix from the interference source to the RIS. Here is the channel matrix from RIS to the receiver. The sender's transmit power, The transmission power of the interference source. Let be the signal-to-interference-plus-noise ratio (SIR) of the received signal at time t-1. Let t be the square wave beamforming vector transmitted at time t. Let be the phase shift of each unit in the RIS at time t. Let be the amplitude change of the RIS units at time t.

4. The communication method based on a reconfigurable smart surface as described in any one of claims 1-3, characterized in that, The step of dynamically updating the reflection coefficient matrix obtained in S4 according to the channel change rate when the adjustment period is reached includes: Pre-correction is performed to address phase deviation caused by Doppler frequency shift; the gain coefficient in the reflection coefficient matrix is ​​dynamically adjusted based on the received signal strength; and the transmit beam actively probes channel state information to address non-cooperative interference.

5. A communication device based on a reconfigurable smart surface, characterized in that, A communication method based on a reconfigurable smart surface according to any one of claims 1-4, comprising: The initial setup module is used to set the initial reflection coefficient matrix of the reconfigurable smart surface RIS and the initial beamforming vector of the transmitter. The first control module is used to control the transmitter to transmit a first communication signal via the first direct link and the first RIS auxiliary link corresponding to the initial reflection coefficient matrix using the initial beamforming vector; the interference source to transmit a first interference signal via the second direct link and the second RIS auxiliary link corresponding to the initial reflection coefficient matrix; and the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the first communication signal and the first interference signal in real time. The iterative optimization module is used to jointly optimize the current beamforming vector and the current RIS reflection parameters of the transmitter based on the channel estimation results acquired in real time using a deep deterministic policy gradient algorithm. The deep deterministic policy gradient algorithm includes: a reward function with the core of maximizing the signal-to-interference-plus-noise ratio of the received signal; a Markov decision process state space containing channel state information, power constraints, and historical performance indicators; and a continuous action space containing the beamforming vector, RIS phase offset, and amplitude gain. The second control module is used to control the transmitter to transmit a second communication signal via the first direct link and the third RIS auxiliary link corresponding to the current reflection coefficient matrix using the initial beamforming vector; the interference source to transmit a second interference signal via the second direct link and the fourth RIS auxiliary link corresponding to the current reflection coefficient matrix; and the receiver to measure the signal-to-interference-plus-noise ratio of the received signals corresponding to the second communication signal and the second interference signal in real time. The iterative judgment module is used to directly switch to the loop optimization module when the adjustment period has not been reached; when the adjustment period is reached, the reflection coefficient matrix output by the second control module is dynamically updated according to the channel change rate, and then the loop optimization module is switched to.

6. A communication system based on a reconfigurable smart surface, characterized in that, include: The method comprises a transmitter, a RIS (Radio Interference Array), an interference source, a receiver, a memory, and a processor. The transmitter sends a communication signal to the receiver via its corresponding RIS auxiliary link and a direct link, while the interference source sends an interference signal to the receiver via its corresponding RIS auxiliary link and a direct link. The receiver measures the signal-to-interference-plus-noise ratio (SIR) of the received signals corresponding to the communication signal and the interference signal in real time. The memory stores a computer program. The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images