Search

US-12627335-B2 - Digital twin-based deduction and optimization method and system for intelligent reflecting surface communication system

US12627335B2US 12627335 B2US12627335 B2US 12627335B2US-12627335-B2

Abstract

Disclosed is a digital twin-based deduction and optimization method and system for an intelligent reflecting surface (IRS) communication system, including: collecting data from a scenario, including relevant data of an IRS physical model and real channel data; performing real-time data transmission on the collected data; establishing a digital twin three-dimensional (3D) model in a digital twin space based on the data after the real-time data transmission; establishing an IRS reflection mechanism model before fusing with the digital twin 3D model, the generalized Snell Equation being used to simplify a complex system in the IRS reflection mechanism model; deducing and optimizing the IRS communication system to obtain an optimization strategy; and feeding the optimization strategy back to the real world to realize the deduction and optimization of the IRS communication system.

Inventors

  • Haixia Zhang
  • Qiaojian Han
  • Dongfeng Yuan

Assignees

  • SHAN DONG UNIVERSITY

Dates

Publication Date
20260512
Application Date
20240722
Priority Date
20231128

Claims (7)

  1. 1 . A digital twin-based deduction and optimization method for an intelligent reflecting surface (IRS) communication system, comprising: collecting data from a scenario, comprising relevant data of an IRS physical model and real channel data; performing real-time data transmission on the collected data; establishing a digital twin three-dimensional (3D) model in a digital twin space based on the data after the real-time data transmission; establishing an IRS reflection mechanism model before fusing with the digital twin 3D model, the generalized Snell Equation being used to simplify a complex system in the IRS reflection mechanism model; deducing and optimizing the IRS communication system to obtain an optimization strategy; and feeding the optimization strategy back to the real world to realize the deduction and optimization of the IRS communication system, wherein the relevant data of the IRS physical model and the real channel data comprise: electromagnetic waves; angles of the electromagnetic waves; a relative distance and position between a base station (BS), the IRS, and user equipment (UE); temperature and humidity; actual communication traffic or dynamic changes of the UE; phase and frequency changes of electromagnetic waves reflected by the IRS; energy or power of the electromagnetic waves transmitted by the BS and reflected by the IRS; an operating state of the IRS; ambient noise; delay, bandwidth, and data transmission rate of an actual network communication; 3D appearance images; and 3D point cloud data; the establishing a digital twin 3D model in a digital twin space based on the data after the real-time data transmission, comprising: performing static content modeling on background information with little change, while performing dynamic content modeling on user information with great change, modeling steps being as follows: data cleaning: processing sensor data, namely, the acquired relevant data of the IRS physical model and real channel data; 3D modeling: creating geometry and appearance of the digital twin 3D model; attribute assignment: assigning physical, chemical, or other relevant attributes to the digital twin 3D model; and model verification: comparing with real equipment or environment to ensure accuracy; the establishing an IRS reflection mechanism model before fusing with the digital twin 3D model comprises: (1) establishment of the IRS reflection mechanism model defining characteristics of IRS units; simulating electromagnetic wave propagation; optimization of algorithm design: dynamically adjusting reflection characteristics of the IRS units to optimize a signal path; and integration of environmental factors: integrating the environmental factors into the IRS reflection mechanism model; (2) establishment of the digital twin 3D model creating a digital twin 3D model of the physical world; integrating sensor data; simulation and analysis: running simulations in the digital twin 3D model to analyze the performance of the IRS and its impact on signal coverage; (3) fusion of the IRS reflection mechanism model and the digital twin 3D model integrating the IRS reflection mechanism model into the digital twin 3D model; dynamic data exchange: ensuring real-time data exchange between the IRS reflection mechanism model and the digital twin 3D model; visualization and analysis: analyzing and demonstrating influences of the IRS and communication performance under different conditions using visualization tools of the digital twin 3D model; real-time updates and iterations: updating the digital twin 3D model in real-time according to real-world changes and IRS performance data, to keep the digital twin 3D model up-to-date and accurate; the generalized Snell Equation being used to simplify a complex system in the IRS reflection mechanism model comprises: (1) transmitting a signal from the BS, that is, transmitting a plurality of beams of light, wherein initialization parameters comprise radial widths of the beams of light, a position of the BS in a 3D geometric space, and the rotation amount of the BS in a 3D space, and specifically comprises directions of the transmitted electromagnetic waves, central position of the IRS, the rotation amount of the IRS in the 3D space, the side lengths, thicknesses, and number of the IRS units, an overall size of an IRS plane, relative position of the IRS units and position thereof in space; (2) solving a normal line of the IRS plane; (3) performing ray detection on the transmitted electromagnetic waves to determine whether the IRS is within the range; detecting if the ray intersects the plane, as shown in Formula (1): t = ( P 0 - L 0 ) · N L · N , ( 1 ) wherein in Formula (1), P 0 is a point on a surface of an object to be detected, L 0 is a starting point of the ray, L is a direction of the ray, N is a normal line to the surface of the object, and “⋅” represents a dot product, t is a scalar, if t is less than 0 or greater than 1, then the ray does not intersect the surface of the object; if t is between 0 and 1, then the ray intersects the surface of the object; calculating an intersection point P by Formula (2): P = L 0 + t * L ; ( 2 ) (4) determining whether the ray is within the range of a plate, namely, whether a coverage range of the ray overlaps with a boundary of the plate; calculating a position of a collision point relative to the IRS plate; (5) finding an index of an IRS unit where a light beam passes through reflection, and acquiring a phase parameter of the index; (6) using the generalized Snell Equation to calculate abnormal reflection to obtain angles of reflection ray θ t and θ i , as shown in Formulas (3) and (4): sin ⁢ ( θ t ) ⁢ n t - sin ⁢ ( θ i ) ⁢ n i = λ o 2 ⁢ π ⁢ d ⁢ Φ d ⁢ x , ( 3 ) θ i = arcsin ⁢ ( sin ⁢ ( θ t ) - λ 0 2 ⁢ π ⁢ n i ⁢ ❘ "\[LeftBracketingBar]" d ⁢ Φ d ⁢ x ❘ "\[RightBracketingBar]" ) , ( 4 ) wherein in Formulas (3) and (4), θ t is an incident angle, n t is an incident spatial dielectric constant, θ i is a reflection angle, n i is a reflection spatial dielectric constant, λ o is wavelength, dΦ is phase mutation, and dx is reflection displacement difference caused by mutation; (7) solving a reflection ray path, and setting a ray detection interface according to the angles and the collision point; and the deducing and optimizing the IRS communication system comprises: a channel model being represented by a receiving end signal y k as shown in Formula (5) in the transmission process of a channel: y k = ( h r , k H ⁢ Θ ⁢ G + h d , k H ) ⁢ ∑ j = 1 K ⁢ w j ⁢ s j + n k , ( 5 ) wherein in Formula (5), k = 1 , … , K , h r , k H is an incident channel, Θ is a matrix composed of IRS phase parameters, G is a reflection channel, h d , k H is a direct channel, H is a conjugate transpose of a matrix operation, d is a direct distance between transmitter and receiver, w j is a j th beamforming parameter of a transmitting end, s j is a j th transmitting signal, K is the number of antennas at the transmitting end, and n k is noise; therefore, Formula (6) is obtained by a channel capacity Formula: SINR k = ❘ "\[LeftBracketingBar]" ( h r , k H ⁢ Θ ⁢ G + h d , k H ) ⁢ w k ❘ "\[RightBracketingBar]" 2 ∑ j ≠ k K ⁢ ❘ "\[LeftBracketingBar]" ( h r , k H ⁢ Θ ⁢ G + h d , k H ) ⁢ w j ❘ "\[RightBracketingBar]" 2 + σ k 2 , ∀ k , ( 6 ) wherein SINR k is a signal to interference plus noise ratio, σ k 2 is a variance of the noise, wk is a power parameter of beamforming; maximizing the minimum SINR k , that is, max{min(SINR k )}, making resource allocation of the IRS communication system reach Pareto optimality.
  2. 2 . The digital twin-based deduction and optimization method for an IRS communication system according to claim 1 , wherein the relevant data of the IRS physical model and the real channel data are acquired through various sensors, various sensors comprising an electromagnetic wave sensor, an angle sensor, a distance/position sensor, a temperature and humidity sensor, a dynamic load sensor, a phase and frequency sensor, an energy or power sensor, a state monitoring sensor, an ambient noise sensor, a network communication quality sensor, an image data sensor, and a point cloud data sensor, wherein the electromagnetic wave sensor is configured to capture the electromagnetic waves transmitted from the BS and the electromagnetic waves reflected by the IRS; the angle sensor is configured to measuring an angle of the electromagnetic wave incident on the IRS and an exit angle of the electromagnetic wave after reflection in real-time; the distance/position sensor is configured to measure the relative distance and position between the BS, the IRS, and the UE; the temperature and humidity sensor is configured to measure temperature and humidity; the dynamic load sensor is configured to monitor the actual communication traffic or the dynamic change of the UE; the phase and frequency sensor is configured to monitor phase and frequency changes of the electromagnetic waves reflected by the IRS; the energy or power sensor is configured to monitor the energy or power of the electromagnetic waves transmitted by the BS and reflected by the IRS; the state monitoring sensor is configured to monitor the operating state of the IRS; the ambient noise sensor is configured to capture ambient noise or interference; the network communication quality sensor is configured to monitor the delay, bandwidth, and data transmission rate of the actual network communication; the image data sensor is configured to collect 3D appearance images and match accurate positioning; and the point cloud data sensor is configured to collect 3D point cloud data.
  3. 3 . The digital twin-based deduction and optimization method for an IRS communication system according to claim 1 , wherein the relevant data of the IRS physical model and the real channel data are acquired through a real-time communication mode; the real-time communication mode comprises mobile communication, satellite communication, a wireless local area network, wired communication, and Bluetooth; the real-time data transmission is performed through communication protocols comprising but not limited to Socket, HTTP/HTTPS, file transfer protocol (FTP), simple mail transfer protocol (SMTP), message queuing telemetry transport (MQTT), constrained application protocol (CoAP), extensible messaging and presence protocol (XMPP), WebSockets, google remote procedure call (gRPC), RESTful application programming interfaces (APIs), advanced message queuing protocol (AMQP), streaming text orientated messaging protocol (STOMP), real-time transport protocol (RTP), transmission control protocol (TCP), user datagram protocol (UDP), Bluetooth, Bluetooth low energy (BLE), Zigbee, Z-Wave, low-rank adaptation (LoRa), near field communication (NFC), Modbus, Profibus, simple object access protocol (SOAP), and data distribution service (DDS).
  4. 4 . The digital twin-based deduction and optimization method for an IRS communication system according to claim 1 , wherein the deducing and optimizing the IRS communication system is realized using deep reinforcement learning (DRL), comprising: (a) definition of environment and state: environment being a problem that is trying to optimize, that is, max{min(SINR k )}; state being a current configuration of a matrix Θ; (b) definition of action space: the action being to modify certain elements in the matrix Θ to generate a new matrix configuration; defining a reward function: a reward function being defined as Formula (7): Reward = min ⁢ ( SINR k ) ; ( 7 ) (c) selection of a DRL algorithm; (d) training of a model: training the model using the selected DRL algorithm to learn what actions to take in each state to maximize rewards; (e) strategy execution: using the model to determine an optimal action to take for a given matrix Θ after the training is complete; (f) evaluation and adjustment: using test data to evaluate performance of the optimization strategy and adjust as needed.
  5. 5 . The digital twin-based deduction and optimization method for an IRS communication system according to claim 1 , wherein the deducing and optimizing the IRS communication system is realized using a swarm intelligence algorithm comprises: (g) encoding of solutions: encoding a matrix Θ to contain an array or list of all the elements; (h) definition of a fitness function: the fitness function y being defined as Formula (8): y = min ⁢ ( SINR k ) ; ( 8 ) (i) trying to find solutions to maximize the fitness function, comprising: initializing population: randomly generating a set of initial Θ; (j) evaluation of initial population: using the fitness function y to evaluate quality of each solution; (k) iterative optimization: for each iteration, updating the solution according to the selected swarm intelligence algorithm, comprising: crossover, mutation, and/or selection of the solutions; (l) termination conditions: continuing the iteration until the termination condition is satisfied; and (m) return to an optimal solution: returning to a solution with an optimal fitness value.
  6. 6 . The digital twin-based deduction and optimization method for an IRS communication system according to claim 4 , wherein the optimization strategy is fed back to the real world through a real-time communication mode.
  7. 7 . A digital twin-based deduction and optimization system for an intelligent reflecting surface (IRS) communication system, comprising: a data collection module, configured to collect data from a scenario, comprising relevant data of an IRS physical model and real channel data; a data transmission module, configured to perform real-time data transmission on the collected data and feed an optimization strategy back to the real world to realize the deduction and optimization of the IRS communication system; a digital twin three-dimensional (3D) model establishment module, configured to establish a digital twin 3D model in a digital twin space based on the data after the real-time data transmission; an IRS reflection mechanism model establishment module, configured to establish an IRS reflection mechanism model before fusing with the digital twin 3D model, the generalized Snell Equation being used to simplify a complex system in the IRS reflection mechanism model; and a deduction-optimization module, configured to deduce and optimize the IRS communication system to obtain an optimization strategy, wherein the relevant data of the IRS physical model and the real channel data comprise: electromagnetic waves; angles of the electromagnetic waves; a relative distance and position between a base station (BS), the IRS, and user equipment (UE); temperature and humidity; actual communication traffic or dynamic changes of the UE; phase and frequency changes of electromagnetic waves reflected by the IRS; energy or power of the electromagnetic waves transmitted by the BS and reflected by the IRS; an operating state of the IRS; ambient noise; delay, bandwidth, and data transmission rate of an actual network communication; 3D appearance images; and 3D point cloud data; the establishing a digital twin 3D model in a digital twin space based on the data after the real-time data transmission is divided into static content modeling and dynamic content modeling, comprising: performing static content modeling on background information with little change, while performing dynamic content modeling on user information with great change, modeling steps being as follows: data cleaning: processing sensor data, namely, the acquired relevant data of the IRS physical model and real channel data; 3D modeling: creating geometry and appearance of the digital twin 3D model; attribute assignment: assigning physical, chemical, or other relevant attributes to the digital twin 3D model; model verification: comparing with real equipment or environment to ensure accuracy; the establishing an IRS reflection mechanism model before fusing with the digital twin 3D model comprises: (1) establishment of the IRS reflection mechanism model defining characteristics of IRS units; simulating electromagnetic wave propagation; optimization of algorithm design: dynamically adjusting reflection characteristics of the IRS units to optimize a signal path; and integration of environmental factors: integrating the environmental factors into the IRS reflection mechanism model; (2) establishment of the digital twin 3D model creating a digital twin 3D model of the physical world; integrating sensor data; simulation and analysis: running simulations in the digital twin 3D model to analyze the performance of the IRS and its impact on signal coverage; (3) fusion of the IRS reflection mechanism model and the digital twin 3D model integrating the IRS reflection mechanism model into the digital twin 3D model; dynamic data exchange: ensuring real-time data exchange between the IRS reflection mechanism model and the digital twin 3D model; visualization and analysis: analyzing and demonstrating influences of the IRS and communication performance under different conditions using visualization tools of the digital twin 3D model; real-time updates and iterations: updating the digital twin 3D model in real-time according to real-world changes and IRS performance data, to keep the digital twin 3D model up-to-date and accurate; the generalized Snell Equation being used to simplify a complex system in the IRS reflection mechanism model comprises: (1) transmitting a signal from the BS, that is, transmitting a plurality of beams of light, wherein initialization parameters comprise radial widths of the beams of light, a position of the BS in a 3D geometric space, and the rotation amount of the BS in a 3D space, and specifically comprises directions of the transmitted electromagnetic waves, central position of the IRS, the rotation amount of the IRS in the 3D space, the side lengths, thicknesses, and number of the IRS units, an overall size of an IRS plane, relative position of the IRS units and position thereof in space; (2) solving a normal line of the IRS plane; (3) performing ray detection on the transmitted electromagnetic waves to determine whether the IRS is within the range; detecting if the ray intersects the plane, as shown in Formula (1): t = ( P 0 - L 0 ) · N L · N , ( 1 ) wherein in Formula (1), P 0 is a point on a surface of an object to be detected, L 0 is a starting point of the ray, L is a direction of the ray, N is a normal line to the surface of the object, and “⋅” represents a dot product, t is a scalar, if t is less than 0 or greater than 1, then the ray does not intersect the surface of the object; if t is between 0 and 1, then the ray intersects the surface of the object; calculating an intersection point P by Formula (2): P = L 0 + t * L , ( 2 ) (4) determining whether the ray is within the range of a plate, namely, whether a coverage range of the ray overlaps with a boundary of the plate; calculating a position of a collision point relative to the IRS plate; (5) finding an index of an IRS unit where a light beam passes through reflection, and acquiring a phase parameter of the index; (6) using the generalized Snell Equation to calculate abnormal reflection to obtain angles of reflection ray θ t and θ i , as shown in Formulas (3) and (4): sin ⁢ ( θ t ) ⁢ n t - sin ⁢ ( θ i ) ⁢ n i = λ o 2 ⁢ π ⁢ d ⁢ Φ d ⁢ x , ( 3 ) θ i = arcsin ⁢ ( sin ⁢ ( θ t ) - λ 0 2 ⁢ π ⁢ n i ⁢ ❘ "\[LeftBracketingBar]" d ⁢ Φ d ⁢ x ❘ "\[RightBracketingBar]" ) , ( 4 ) wherein in Formulas (3) and (4), θ t is an incident angle, n t is an incident spatial dielectric constant, θ i is a reflection angle, n i is a reflection spatial dielectric constant, λ o is wavelength, dΦ is phase mutation, and dx is reflection displacement difference caused by mutation; (7) solving a reflection ray path, and setting a ray detection interface according to the angles and the collision point; and the deducing and optimizing the IRS communication system comprises: a channel model being represented by a receiving end signal y k as shown in Formula (5) in the transmission process of a channel: y k = ( h r , k H ⁢ Θ ⁢ G + h d , k H ) ⁢ ∑ j = 1 K ⁢ w j ⁢ s j + n k , ( 5 ) wherein in Formula (5), k = 1 , … , K , h r , k H is an incident channel, Θ is a matrix composed of IRS phase parameters, G is a reflection channel, h d , k H is a direct channel, w j is a j th beamforming parameter of a transmitting end, s j is a j th transmitting signal, K is the number of antennas at the transmitting end, and n k is noise; therefore, Formula (6) is obtained by a channel capacity Formula: SINR k = ❘ "\[LeftBracketingBar]" ( h r , k H ⁢ Θ ⁢ G + h d , k H ) ⁢ w k ❘ "\[RightBracketingBar]" 2 ∑ j ≠ k K ⁢ ❘ "\[LeftBracketingBar]" ( h r , k H ⁢ Θ ⁢ G + h d , k H ) ⁢ w j ❘ "\[RightBracketingBar]" 2 + σ k 2 , ∀ k , ( 6 ) wherein SINR k is a signal to interference plus noise ratio, σ k 2 is a variance of the noise; maximizing the minimum SINR k , that is, max{min(SINR k )}, making resource allocation of the IRS communication system reach Pareto optimality.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to Chinese Patent Application Ser. Nos. CN2023116033540 and CN2024100120219 filed on Nov. 28, 2023 and Jan. 4, 2024, respectively. FIELD OF THE INVENTION The present disclosure relates to a digital twin-based deduction and optimization method and system for an intelligent reflecting surface (IRS) communication system in the technical field of digital twin. BACKGROUND OF THE INVENTION With the continuous progress of communication technology, effectively enhancing communication quality and coverage range has become a research hotspot. To cope with this challenge, digital twin, as a new generation of information technology, has begun to receive widespread attention. This technology maps things from the real world to the virtual world, thus providing a new perspective to understand and optimize various scenarios in reality. In the field of communications, digital twin technology has great potential. By combining physical mechanism systems or artificial intelligence algorithms with digital twins, predictions can be made based on existing or historical data, to preview and optimize the communication environment before actual operation. Such a preview simulation contributes to optimizing system parameters and provides feedback on achieving an optimal communication effect in a specific environment. IRS, as a hotspot of communication research in recent years, is particularly suitable for modeling and optimization using digital twins. The IRS consists of low-cost passive reflective elements capable of creating a direct line of sight (LOS) links and accurately delivering communication signals. This is particularly valuable in urban environments because buildings and other obstacles tend to hinder the propagation of signals. However, although the theoretical research of the IRS is deepening, its deployment in actual scenarios is still limited by high cost and high risk. Current research mainly focuses on optimizing communication rates and balancing user communication resources. For example, Chinese patent document CN116667902A discloses “Active Intelligent Reflecting Surface Auxiliary Communication System Mode Selection Method”, and Chinese patent document CN116648861A discloses “System and Method for Using Intelligent Reflecting Surface in Communication System”. Meanwhile, some related technologies have applied 5G network slice and digital twin management terminals. However, there are still some problems and disadvantages with this system. First, the concept of server digital twins is marginalized. For example, Chinese patent document CN114650545A discloses “Beam Parameter Determination Method and Apparatus, and Network Device”, including generating the IRS digital twin of a first IRS panel; and obtaining beam parameters from a network device to the IRS panel and/or form the IRS panel to a terminal according to the digital twin. Chinese patent document CN114928893A discloses “Architecture and Task Unloading Method Based on Intelligent Reflecting Surface”, in which the concept of digital twin optimization is applied. However, the concept of digital twins in the existing patents is not clear enough, and there is no complete step to construct a digital twin system. However, few studies have focused on the specific deployment of IRS in real environments, especially the lack of geometric features of electromagnetic wave reflection caused by the metamaterial abnormal reflection mechanism of IRS. This leads to a certain gap between theoretical research and actual deployment. SUMMARY OF THE INVENTION Given the shortcomings of the prior art, the present disclosure constructs an IRS-oriented three-dimensional (3D) digital twin system and proposes a full-flow mechanism-level preview method. Through the method, the full-flow and full-dimensional preview of the communication process may be realized, and the IRS user optimization method based on swarm intelligence algorithm or deep reinforcement learning (DRL) is established. To ensure real-time information transmission between the virtual and real space, the system further integrates real-time communication technology. In conclusion, combined with the application of digital twin technology and IRS in communication, it provides a new, efficient, and practical optimization framework for future communication systems. Explanation of Terms: 1. Deep Q-Network (DQN) is an algorithm that combines deep learning (DL) and reinforcement learning (RL). It uses a depth neural network to approximate a Q function that estimates the expected return to perform some action in a given state. By combining the expressive learning ability of DL and the decision-making ability of Q-learning, DQN achieves effective learning in a complex environment.2. Proximal strategy optimization (PPO) is a method for training RL algorithms, especially in the family of strategy gradients. It aims to solve some stability and efficiency problems in t