Graph neural network-based intelligent reflecting surface auxiliary vehicle sensing integrated joint beam forming method

Abstract

The invention relates to a graph neural network-based intelligent reflecting surface auxiliary vehicle sensing integrated joint beam forming method, and belongs to the field of Internet of Vehicles. The method aims at maximizing the weighting and signal-to-noise ratio of communication vehicles and sensing vehicles, considers the reflection unit mode constraint of the intelligent reflection surface and the transmission power constraint of the base station, and solves the problem of joint design of active beam forming and passive beam forming in the intelligent reflection surface assisted Internet of Vehicles communication and sensing integrated system. Interaction among the vehicles, the base stations and the intelligent reflecting surfaces is captured through the graph neural network architecture, the influence of the number of the vehicles on the network is ignored, the network applicability is improved, arrangement equivalence is achieved, and the network robustness is enhanced. According to the method, the performance in the aspect of improving weighting and the signal-to-noise ratio by using the channel coefficient is better, Doppler frequency shift and multiple-input-multiple-output in the scene of the Internet of Vehicles are considered at the same time, joint beamforming optimization is performed on communication and target sensing, and the system performance is improved.

Description

Technical Field

[0001] The present invention belongs to the field of vehicle networking and relates to a graph neural network-based intelligent reflective surface-assisted vehicle synaesthesia integrated joint beamforming method. Background Art

[0002] Vehicular communications, also known as the Internet of Vehicles, are a key enabler for autonomous driving, intelligent transportation systems, and in-vehicle infotainment services. As one of the key platforms for the future sixth-generation wireless communication network, the Internet of Vehicles can facilitate the transition from traditional mode vehicles to autonomous vehicles. This makes vehicles gradually become the basic platform for future intelligent transportation systems, integrating vehicles into existing applications rather than just means of transportation. The realization of future autonomous vehicles requires robust connectivity and high service quality between vehicles and everything. Communication-sensing integration and reconfigurable smart surfaces have been considered as two potential enabling technologies for 6G.

[0003] Synaesthesia integration refers to an information processing scheme that realizes the coordination of communication and radar perception. Compared with the system with separate perception and communication functions, synaesthesia integration realizes perception and communication functions on the same hardware and frequency band at the same time, which reduces power consumption and deployment complexity, and significantly improves spectrum efficiency and hardware efficiency. The base station can perceive and communicate with the vehicle at the same time, so the communication between the telematics-assisted telematics-assisted vehicles can utilize the integration gain and coordination gain of synaesthesia integration to obtain more accurate perception information and better wireless communication quality, which will strongly support various intelligent services and improve the performance of vehicles on the road. The millimeter-wave band (mmWave) is an ideal band for the synaesthesia-integrated system, which can provide huge bandwidth to improve the transmission rate and perception accuracy of the synaesthesia-integrated system. However, the path loss in the millimeter-wave band is very serious, which will weaken the signal and create blind spots in the vehicle network, thereby reducing the perception and communication performance. To solve this problem, smart reflective surfaces can be deployed in the telematics-assisted vehicle link, which can ensure the quality of the wireless link in a dynamic and uncertain environment.

[0004] The smart reflective surface is a two-dimensional structure composed of many low-cost reconfigurable elements, whose phase can be adjusted by using an intelligent controller connected to the smart reflective surface. It can passively reflect the received base station transmission signal and control the reflection direction of the signal, so as to establish a strong virtual line-of-sight link between the base station and the vehicle, improve the quality of the communication link and the perception link, without any radio frequency hardware and additional energy consumption, which is a good choice for the interawareness integration system. Smart reflective surfaces can be attached to existing infrastructure, which helps them to be deployed in wireless environments and installed on building exterior walls, road signs and other places, greatly reducing the installation cost. Smart reflective surfaces can achieve directional signal enhancement and extend the transmission range, which provides solid support for the Internet of Vehicles with high data rates and wide coverage requirements, while providing extremely low interruption probability for remote driving. The above advantages of smart reflective surfaces make smart reflective surface-assisted vehicle communication an attractive choice for enhancing vehicle network connection. The present invention aims at the beamforming problem of the Internet of Vehicles link in the vehicle communication and perception integration scenario. Due to the high fading characteristics of the mmWave frequency band, the communication between the vehicle and the base station is assisted by introducing a reconfigurable smart reflective surface to improve the link quality and stability. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide an intelligent reflective surface-assisted vehicle synaesthesia integrated joint beamforming method based on graph neural network, which solves the limitations of the prior art when facing vehicle network beamforming.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A graph neural network-based intelligent reflective surface-assisted vehicle synaesthesia integrated joint beamforming method, the method comprising the following steps:

[0008] S1. Based on the telematics integration system of the Internet of Vehicles, a first channel model between the telematics integration device, the intelligent reflective surface and the communication vehicle and a second channel model between the telematics integration device, the intelligent reflective surface and the perception vehicle are constructed respectively;

[0009] S2. Obtaining a communication signal-to-noise ratio and a perception signal-to-noise ratio of the vehicle networking interaural integration system based on the first channel model and the second channel model;

[0010] S3, determining a communication signal-to-noise ratio and a perception signal-to-noise ratio based on the communication signal and the perception signal, with the goal of maximizing the weighted sum signal-to-noise ratio of the communication vehicle and the perception vehicle;

[0011] S4. Build a graph neural network beamforming architecture, which effectively solves the joint design problem of active / passive beamforming by capturing the interaction between vehicles, base stations, and smart reflective surfaces. Use the graph neural network beamforming architecture to learn the direct mapping relationship between channel coefficients and optimal parameter settings.

[0012] Furthermore, in S1, a graph neural network-based intelligent reflective surface-assisted vehicle synaesthesia integrated joint beamforming method is characterized in that: in S1, a first channel model between a synaesthesia integrated device, an intelligent reflective surface and a communication vehicle and a second channel model between a synaesthesia integrated device, an intelligent reflective surface and a perception vehicle are constructed respectively:

[0013] The first channel model is: the receiving signal of the communication vehicle;

[0014] The received signal of the communication vehicle is the sum of two parts. The first part is the product of the base station beamforming matrix and the base station-RIS-communication vehicle channel, and the second part is additive Gaussian white noise.

[0015] The second channel model is: the received signal of the sensing vehicle;

[0016] The received signal of the sensing vehicle is the sum of two parts. The first part is the product of the base station beamforming matrix and the base station-RIS-sensing vehicle channel, and the second part is additive Gaussian white noise.

[0017] Furthermore, in S2, the first channel model and the second channel model obtain the transmission signal and the perception signal of the vehicle networking interawareness integration system:

[0018] The signal-to-noise ratio of communication vehicle k is defined as the ratio of the communication vehicle channel gain to the additive white Gaussian noise;

[0019] Therefore, the total signal-to-noise ratio of the communicating vehicles is: the sum of the vehicle signal-to-noise ratios;

[0020] The radar signal-to-noise ratio of the mth sensing vehicle is defined as: the ratio of the sensing vehicle channel gain to the additive white Gaussian noise;

[0021] Therefore, the total signal-to-noise ratio of the perceived vehicle is: the sum of the perceived vehicle signal-to-noise ratios.

[0022] Furthermore, in S3, the communication signal-to-noise ratio and the perception signal-to-noise ratio are determined based on the communication signal and the perception signal, with the goal of maximizing the weighted sum signal-to-noise ratio of the communication vehicle and the perception vehicle:

[0023] defining a weighted sum of maximizing communication signal-to-noise ratio and perception signal-to-noise ratio with respect to active beamforming and passive beamforming;

[0024] Furthermore, in S4, a graph neural network beamforming architecture is constructed. The architecture effectively solves the joint design problem of active / passive beamforming by capturing the interaction between the vehicle, the base station, and the intelligent reflective surface. Specifically,

[0025] The graph consists of K+M+2 nodes, specifically, K communication vehicles are represented by nodes 1 to K, M target vehicles are represented by nodes K+1 to K+M, and the smart reflective surface and base station are represented by nodes K+M+1 and K+M+2, respectively. The representation vector of a node should contain all the useful information required to establish the mapping. The representation vector is updated in a layer-by-layer manner, where each layer consists of combination and aggregation operations that take the representation vectors in the previous layer as input. After the update, all representation vectors have enough information to optimize the design of active beamforming for base stations and passive beamforming for smart reflective surfaces. The optimal settings can be obtained directly from the communication vehicle representation vector, the target vehicle representation vector, the smart reflective surface representation vector, and the base station representation vector.

[0026] By feeding the channel coefficients into a well-trained graph neural network, the optimized base station active beamforming and intelligent reflector passive beamforming designs are obtained. The proposed graph neural network-based learning algorithm consists of an initialization layer, multiple update layers, and a parameter generation layer. First, the channel coefficients are transformed into an initial representation vector through the initialization layer, then the representation vectors of all nodes are updated through D update layers, and finally the base station active beamforming matrix and RIS passive beamforming matrix are obtained through the parameter generation layer.

[0027] The beneficial effects of the present invention are:

[0028] First, the present invention realizes the simultaneous communication of the base station and target perception in the interawareness integrated system based on the intelligent reflective surface. By designing the active beamforming of the base station and the passive beamforming of the intelligent reflective surface, the weighted sum of the signal-to-noise ratio of all communicating vehicles and sensing vehicles is optimized, thereby improving the overall performance of the system.

[0029] Second, the present invention takes Doppler frequency shift and MIMO technology into consideration, making it more suitable for Internet of Vehicles scenarios and enhancing the adaptability and reliability of the system.

[0030] Third, the present invention proposes a graph neural network beamforming architecture, which directly maps the channel coefficients to the base station active beamforming and the intelligent reflector passive beamforming, reducing the computational complexity and improving the model aggregation accuracy.

[0031] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0033] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0034] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0035] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0036] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0037] See also Figure 1 , which is a flow chart of the intelligent reflective surface-assisted vehicle synaesthesia integrated joint beamforming method based on graph neural network.

[0038] The graph neural network-based intelligent reflective surface-assisted vehicle synaesthesia integrated joint beamforming method of the present invention comprises the following steps:

[0039] S1. Initialization

[0040] The first channel model between the synaesthesia integrated device, the intelligent reflective surface and the communication vehicle and the second channel model between the synaesthesia integrated device, the intelligent reflective surface and the perception vehicle are constructed respectively. Since the direct channel, that is, the base station-communication vehicle link is severely blocked by obstacles, the direct channel is ignored and only the intelligent reflective surface auxiliary channel, that is, the base station-intelligent reflective surface-communication vehicle link, is considered. The duration of the transmission frame is T, and the transmission frame is divided into U subframes, and the duration of each subframe is recorded as T / U. The position of the vehicle is unlikely to change significantly within a subframe, so it is assumed that the vehicle position is unchanged and fixed during a subframe, but it will change from one subframe to another. Therefore, the first channel model is: the received signal of the communication vehicle; wherein, the received signal of the communication vehicle is the sum of two parts. The first part is the product of the base station beamforming matrix and the base station-RIS-communication vehicle channel, and the second part is additive Gaussian white noise. The second channel model is: the received signal of the perception vehicle; wherein, the received signal of the perception vehicle is the sum of two parts. The first part is the product of the base station beamforming matrix and the base station-RIS-sensing vehicle channel, and the second part is additive white Gaussian noise.

[0041] S2. Calculate the communication signal-to-noise ratio and the perception signal-to-noise ratio

[0042] The signal-to-noise ratio of the communication vehicle and the signal-to-noise ratio of the perception vehicle are obtained based on the first channel model and the second channel model. The signal-to-noise ratio of the communication vehicle k is: the ratio of the communication vehicle channel gain to the additive white Gaussian noise; therefore, the total signal-to-noise ratio of the communication vehicle is: the sum of the vehicle signal-to-noise ratios; the radar signal-to-noise ratio of the mth perception vehicle is: the ratio of the perception vehicle channel gain to the additive white Gaussian noise; therefore, the total signal-to-noise ratio of the perception vehicle is: the sum of the perception vehicle signal-to-noise ratios.

[0043] S3. Construct optimization goals

[0044] Determine the communication signal-to-noise ratio and the perception signal-to-noise ratio based on the communication signal and the perception signal, with maximizing the weighted sum of the communication vehicle and the perception vehicle signal-to-noise ratio as the optimization goal;

[0045] S4. Constructing the beamforming architecture of graph neural network

[0046] A graph neural network is constructed to learn a direct mapping between channel coefficients and optimal parameter settings. The mapping function is denoted as κ(·), which maps channel coefficients (i.e., base station-RIS, RIS-communication vehicle, and RIS-sensing vehicle) to base station transmit beamforming and RIS passive beamforming. Therefore, solving the problem can be interpreted as learning the optimal mapping function. The properties of deep neural networks are used to parameterize the mapping function κ(·) between channel coefficients and optimized parameter settings, and then the neural network is trained in a data-driven manner to learn the optimal mapping function.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for joint beamforming based on graph neural network with intelligent reflective surface assisted vehicle synaesthesia integration, characterized by: The method comprises the following steps: S1. Based on the telematics integration system of the Internet of Vehicles, a first channel model between the telematics integration device, the intelligent reflective surface and the communication vehicle and a second channel model between the telematics integration device, the intelligent reflective surface and the perception vehicle are respectively constructed; S2. Obtaining a communication signal-to-noise ratio and a perception signal-to-noise ratio of the vehicle networking interaural integration system based on the first channel model and the second channel model; S3, determining a communication signal-to-noise ratio and a perception signal-to-noise ratio based on the communication signal and the perception signal, with the goal of maximizing the weighted sum signal-to-noise ratio of the communication vehicle and the perception vehicle; S4. Construct a graph neural network beamforming architecture, which solves the joint design problem of active / passive beamforming by capturing the interaction between vehicles, base stations, and smart reflective surfaces; and adopts a graph neural network beamforming architecture to learn the direct mapping relationship between channel coefficients and optimal parameter settings.

2. According to claim 1, a graph neural network-based intelligent reflective surface-assisted vehicle synaesthesia integrated joint beamforming method is characterized by: In S1, a first channel model between the synaesthesia integrated device, the intelligent reflective surface and the communication vehicle and a second channel model between the synaesthesia integrated device, the intelligent reflective surface and the perception vehicle are constructed respectively: The first channel model is: the receiving signal of the communication vehicle; Among them, the received signal of the communication vehicle is the sum of two parts; the first part is the product of the base station beamforming matrix and the base station-RIS-communication vehicle channel, and the second part is additive Gaussian white noise; The second channel model is: the received signal of the sensing vehicle; Among them, the received signal of the sensing vehicle is the sum of two parts; the first part is the product of the base station beamforming matrix and the base station-RIS-sensing vehicle channel, and the second part is additive white Gaussian noise.

3. The method for integrating vehicle synaesthesia with intelligent reflective surface based on graph neural network according to claim 2 is characterized by: In S2, the first channel model and the second channel model obtain the transmission signal and the perception signal of the vehicle network interawareness integration system; The signal-to-noise ratio of communication vehicle k is defined as the ratio of the communication vehicle channel gain to the additive white Gaussian noise; The total signal-to-noise ratio of the communicating vehicles is: the sum of the vehicle signal-to-noise ratios; The radar signal-to-noise ratio of the mth sensing vehicle is defined as: the ratio of the sensing vehicle channel gain to the additive white Gaussian noise; The total signal-to-noise ratio of the perceived vehicle is: the sum of the perceived vehicle signal-to-noise ratios.

4. The method for joint beamforming based on graph neural network and intelligent reflective surface assisted vehicle synaesthesia integration according to claim 3 is characterized by: In S3, the communication signal-to-noise ratio and the perception signal-to-noise ratio are determined based on the communication signal and the perception signal, with the goal of maximizing the weighted sum signal-to-noise ratio of the communication vehicle and the perception vehicle: Define the weighted sum of maximizing the communication signal-to-noise ratio and the perception signal-to-noise ratio with respect to the active beamforming W and the passive beamforming Θ.

5. The method for joint beamforming based on graph neural network and intelligent reflective surface assisted vehicle synaesthesia integration according to claim 4 is characterized by: In S4, a graph neural network beamforming architecture is constructed. This architecture effectively solves the joint design problem of active / passive beamforming by capturing the interaction between vehicles, base stations, and smart reflective surfaces. Specifically: The graph consists of K+M+2 nodes, K communication vehicles are represented by nodes 1 to K, M target vehicles are represented by nodes K+1 to K+M, and the smart reflective surface and base station are represented by nodes K+M+1 and K+M+2, respectively; the representation vector of the node should contain all the useful information required to establish the mapping; the representation vector is updated in a layer-by-layer manner, where each layer consists of combination and aggregation operations that take the representation vectors in the previous layer as input; after the update, all the representation vectors have enough information to optimize the design of active beamforming of the base station and passive beamforming of the smart reflective surface; the optimal setting is obtained from the communication vehicle representation vector, the target vehicle representation vector, the smart reflective surface representation vector and the base station representation vector; By feeding the channel coefficients into a well-trained graph neural network, the optimized base station active beamforming and smart reflector passive beamforming designs are obtained; the proposed graph neural network-based learning algorithm consists of an initialization layer, multiple update layers, and a parameter generation layer; first, the channel coefficients are transformed into representation vectors through the initialization layer, and then the representation vectors of all nodes are updated through D update layers, and finally the base station active beamforming matrix and RIS passive beamforming matrix are obtained through the parameter generation layer.

Images