Car-road cooperation communication method, device, equipment and medium

Abstract

The application relates to a vehicle-road cooperation communication method, device, equipment and medium. The method comprises the following steps: determining a maximum data successful transmission probability, determining an optimal semantic compression ratio based on an information available probability model and the maximum data successful transmission probability, extracting semantic features of traffic scene image data according to the optimal semantic compression ratio based on an ISSC encoder in a preset image segmentation semantic communication model to generate a semantic graph corresponding to the traffic scene image data, and pushing the semantic graph corresponding to the traffic scene image data to a vehicle-mounted unit. An ISSC decoder in the vehicle-mounted unit decodes the received semantic graph corresponding to the traffic scene image data, reconstructs image segmentation determined image semantic segmentation data, and completes vehicle-road cooperation communication. The application can solve the problems of large data transmission and high bandwidth resource consumption in vehicle-road cooperation.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation, and in particular to a vehicle-road cooperative communication method, corresponding device, electronic device and computer-readable storage medium. Background Technology

[0002] Vehicle-to-infrastructure (V2I) communication is an application based on advanced wireless communication and integrating next-generation Internet technologies, aiming to fully realize real-time information interaction between vehicles and between vehicles and the road network. By collecting and fusing dynamic traffic information in vehicle traffic scenarios, V2I enables vehicles to obtain various traffic conditions in real time, assisting in autonomous driving or other safety control operations, as well as road collaborative management, thereby achieving efficient collaboration between people, vehicles, and roads. Its main goals are to improve traffic safety, enhance traffic efficiency, and create a safe and efficient intelligent transportation system that serves society. Currently, a large amount of research on V2I applications has been carried out globally. V2I scenarios mainly include roadside units (RSUs) and on-board units (OBUs). Roadside units are used for V2I (Vehicle-to-Infrastructure) communication and are key devices for realizing intelligent road and vehicle-road collaboration. They are installed on the roadside and communicate and exchange data bidirectionally with nearby passing vehicles. In vehicle-road cooperative scenarios, the Onboard Unit (OBU) needs to acquire real-time road condition information, including scene and weather data. By deploying AIGC (Artificial Intelligence Generated Content) and other AI-generated content technologies on the OBU, the received semantic data can be linked with semantic graphs using AIGC multimodal technology to obtain realistic scene maps, thus assisting in autonomous driving. The Roadside Unit (RSU) can connect to existing roadside electronic devices, such as high-definition cameras, to acquire relevant data and upload it to the cloud or distribute it to vehicles—a crucial role of the RSU, especially on highways. Highway RSUs integrate perception units and connect to devices like high-definition cameras, enabling them to detect and identify abnormal road conditions—such as pedestrians or animals entering the road, vehicles stopping in the emergency lane, vehicles driving in the wrong direction, vehicle breakdowns, and traffic accidents. This collected image information can then be uploaded to the cloud in real time and simultaneously disseminated in advance to vehicles equipped with onboard terminals that are about to approach abnormal road sections. However, the large amount of image data interaction between the RSU and OBU is very time-consuming and puts a burden on the network. Therefore, semantic communication is introduced to extract semantic data from the raw data and express the desired meaning, filtering out irrelevant information from the original image to reduce information redundancy. After receiving semantic data from the RSU, the on-board unit provides various services through artificial intelligence content generation technologies such as AIGC.

[0003] Currently, data transmission stability and efficiency in existing vehicle-to-infrastructure (V2I) systems are poor. In existing V2I scenarios, data acquired by the roadside unit (RSU) is transmitted to the on-board unit (OBU) without processing, resulting in a large amount of data transmission. This consumes more bandwidth resources or increases transmission latency. However, due to spectrum resource limitations and complex communication conditions in the communication environment, it is difficult to maintain a reliable connection to transmit large amounts of data, and the increased latency is detrimental to the real-time performance of data sharing.

[0004] Existing semantic communication applications do not consider the semantic compression ratio issue under data AOI constraints. In vehicle-to-everything (V2X) scenarios, using a higher compression ratio requires fewer channel resources but leads to semantic distortion, while using a lower compression ratio requires more channel resources. Therefore, data may expire and become invalid because the data AOI does not meet the constraints.

[0005] Traditional image segmentation semantic communication models suffer from weak feature extraction capabilities and high computational complexity. In existing vehicle-to-everything (V2X) systems, most image segmentation semantic communication models used for preprocessing transmitted data are convolutional neural networks (CNNs). However, the semantic extraction capability of these networks is limited by the size of the convolutional kernels, resulting in weak feature extraction and high computational complexity.

[0006] In summary, existing technologies for vehicle-road cooperative systems suffer from poor data transmission stability, low transmission efficiency, and large data volume. They also fail to consider the semantic compression ratio under data AOI constraints, and traditional image segmentation semantic communication models have weak feature extraction capabilities and high computational complexity. To address these issues, the applicant has made corresponding explorations. Summary of the Invention

[0007] The purpose of this application is to solve the above-mentioned problems by providing a vehicle-road cooperative communication method, corresponding device, electronic device and computer-readable storage medium.

[0008] To achieve the various objectives of this application, the following technical solution is adopted:

[0009] A vehicle-road cooperative communication method proposed to meet one of the purposes of this application includes:

[0010] Acquire traffic scene image data, vehicle historical trajectory data, and vehicle driving data, including vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle;

[0011] A preset vehicle position prediction model is used to determine the vehicle's position coordinates in the next period based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle. Based on the vehicle position coordinates, the communication distance and channel resources between the roadside unit and the vehicle-mounted unit are determined.

[0012] Based on the communication distance and channel resources, determine the probability of successful data transmission and the probability of successful task execution under successful data transmission, and construct an information availability probability model based on the probability of successful data transmission and the probability of successful task execution under successful data transmission.

[0013] Determine the maximum probability of successful data transmission, determine the optimal semantic compression ratio based on the information availability probability model and the maximum probability of successful data transmission, and extract the semantic features of the traffic scene image data based on the optimal semantic compression ratio using the ISSC encoder in the preset image segmentation semantic communication model to generate a semantic map corresponding to the traffic scene image data.

[0014] The semantic map corresponding to the traffic scene image data is pushed to the vehicle unit. The ISSC decoder in the vehicle unit decodes the received semantic map corresponding to the traffic scene image data, reconstructs the image segmentation, and determines the data after image semantic segmentation to complete vehicle-road cooperative communication.

[0015] Optionally, the step of determining the vehicle's position coordinates in the next period using a preset vehicle position prediction model based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and the vehicle's driving angle includes:

[0016] The vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle are input into a preset vehicle position prediction model, which is a CNN-GRU prediction model.

[0017] In the CNN network structure, the CNN network structure includes convolutional layers, ReLU activation function layers, max pooling layers, and flattening layers. The convolutional layer uses a filter to generate a feature matrix based on the vehicle's historical trajectory data, the vehicle's speed, acceleration, distance to the left lane, distance to the right lane, and the vehicle's driving angle. The ReLU activation function layer is used to calculate the output of the feature matrix. The max pooling layer is used to reduce the size of the feature matrix. The flattening layer is used to flatten the output of the previous layer.

[0018] The features extracted by the CNN network structure are passed to the GRU model to predict the vehicle's position coordinates in the next period.

[0019] Optionally, the step of determining the probability of successful data transmission and the probability of successful task execution under successful data transmission based on the communication distance and channel resources, and constructing an information availability probability model based on the probability of successful data transmission and the probability of successful task execution under successful data transmission, includes:

[0020] The information can be used in a probability model as follows:

[0021] Φ=η(o)×P(a≤a0)

[0022] Where P(a≤a0) represents the probability of successful data transmission, a0 is the upper limit of the data AOI, η(ο) represents the probability of successful task execution under successful transmission, ο is the semantic compression ratio, and the data AOI is the sum of semantic compression processing delay, data transmission delay, and data propagation delay.

[0023] Optionally, the steps for determining data transmission delay include:

[0024] Determine the data volume and data transmission rate of traffic scene image data after semantic compression;

[0025] The transmission delay is determined based on the ratio between the amount of semantically compressed traffic scene image data and the data transmission rate.

[0026] Optionally, the steps for determining data propagation delay include:

[0027] Determine the communication distance and data propagation rate between the roadside unit and the vehicle-mounted unit;

[0028] The data propagation delay is determined by calculating the ratio between the communication distance between the roadside unit and the vehicle-mounted unit and the data propagation rate.

[0029] Optionally, the step of the ISSC encoder in the preset image segmentation semantic communication model extracting semantic features of the traffic scene image data according to the optimal semantic compression ratio to generate a semantic map corresponding to the traffic scene image data includes:

[0030] During the encoding process, the ISSC encoder extracts and aggregates semantic features from traffic scene image data according to the optimal semantic compression ratio through a semantic feature extractor and a semantic feature aggregator. The semantic feature extractor is a Swin transformer network.

[0031] The Swin transformer network consists of four stages, each with an even number of Swin transformer network blocks. The Swin transformer network blocks divide the input image into multiple independent small regions to determine window partitions, and calculate self-attention through a multi-head self-attention mechanism module based on window partitions.

[0032] Each rule-based Swin transformer network block follows a multi-head self-attention mechanism module based on shifted window partitions, which shifts the window partitions. Pixels enable interaction between adjacent pixels in different window partitions to determine feature F. sc ;

[0033] The feature F sc Feature fusion is performed after a fully connected layer. The compression ratio of the ISSC encoder is adjusted by the number of neurons in the fully connected layer to determine the output size of the semantic graph corresponding to the traffic scene image data.

[0034] Optionally, the step of the ISSC decoder in the vehicle unit decoding the semantic map corresponding to the received traffic scene image data and reconstructing image segmentation to determine the semantically segmented data includes:

[0035] The traffic scene image data includes one or more of the following: traffic flow image data, roadblock image data, pedestrian crossing image data, and traffic light image data.

[0036] The ISSC decoder consists of a semantic feature decoder and a reconstructor;

[0037] The semantic feature decoder consists of a fully connected layer and two convolutional layers, and the reconstructor consists of two convolutional layers, three upsampling layers, a normalized activation layer, and an ArgMax layer.

[0038] A vehicle-road cooperative communication device provided for another purpose of this application includes:

[0039] The data acquisition module is configured to acquire traffic scene image data, vehicle historical trajectory data, and vehicle driving data. The vehicle driving data includes vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle.

[0040] The communication resource acquisition module is configured to use a preset vehicle position prediction model to determine the vehicle position coordinates in the next period based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle, and to determine the communication distance and channel resources between the roadside unit and the vehicle-mounted unit based on the vehicle position coordinates.

[0041] The information availability model construction module is configured to determine the probability of successful data transmission and the probability of successful task execution under successful data transmission based on the communication distance and channel resources, and to construct an information availability probability model based on the probability of successful data transmission and the probability of successful task execution under successful data transmission.

[0042] The semantic graph generation module is configured to determine the maximum probability of successful data transmission, determine the optimal semantic compression ratio based on the information availability probability model and the maximum probability of successful data transmission, and extract the semantic features of the traffic scene image data based on the optimal semantic compression ratio using the ISSC encoder in the preset image segmentation semantic communication model to generate a semantic graph corresponding to the traffic scene image data.

[0043] The semantic graph push module is configured to push the semantic graph corresponding to the traffic scene image data to the vehicle unit. The ISSC decoder in the vehicle unit decodes the received semantic graph corresponding to the traffic scene image data, reconstructs the image segmentation, and determines the data after image semantic segmentation to complete vehicle-road cooperative communication.

[0044] An electronic device provided for another purpose of this application includes a central processing unit and a memory, the central processing unit being configured to invoke and run a computer program stored in the memory to perform the steps of the vehicle-road cooperative communication method of this application.

[0045] A computer-readable storage medium is provided for another purpose of this application, which stores, in the form of computer-readable instructions, a computer program implemented according to the vehicle-road cooperative communication method, which, when invoked by a computer, executes the steps included in the corresponding method.

[0046] Compared to existing technologies, this application addresses the problems of poor data transmission stability, low transmission efficiency, and large data volume in vehicle-to-everything (V2X) communication, the lack of consideration for semantic compression ratio under data AOI constraints, and the weak feature extraction capability and high computational complexity of traditional image segmentation semantic communication models. This application offers the following advantages, including but not limited to:

[0047] Firstly, this application uses an image segmentation semantic communication model to process the pre-transmission data of vehicles, reducing the amount of data while ensuring data characteristics, thereby reducing the bandwidth resources consumed in the transmission process and maintaining a fast and reliable data sharing connection, so as to solve the problems of large data volume, communication cost and high latency in vehicle-road cooperation.

[0048] Secondly, the image segmentation semantic communication model of this application can guarantee strong feature extraction and low computational complexity, and achieve reliable fidelity before and after data transmission, so as to solve the problem of large data volume and high bandwidth consumption when transmitting data in vehicle-road cooperative systems.

[0049] Third, this application introduces a lightweight prediction model, CNN-GRU, into the roadside unit to predict the vehicle's position in the next period, further obtaining the vehicle's communication distance and channel resources. It also dynamically adjusts the semantic compression ratio by combining data AOI and data fidelity to maximize the success rate of information availability.

[0050] Furthermore, the image segmentation semantic communication system ISSC proposed in this application can solve the problems of large data volume and high bandwidth consumption when roadside units transmit data in vehicle-road cooperative systems, while ensuring strong feature extraction and low computational complexity, ensuring data fidelity, and reducing compression latency.

[0051] This application introduces a lightweight prediction model, CNN-GRU, into the roadside unit (RSU) to predict the next-period vehicle position, enabling more accurate tracking of real-time vehicle information. By obtaining vehicle communication status through location information, the application further models the data AOI, analyzes the relationship between data AOI, fidelity, and semantic compression ratio, and maximizes the success rate of information availability by dynamically adjusting the semantic compression ratio, which has practical application significance. Attached Figure Description

[0052] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0053] Figure 1 This is a schematic diagram of the vehicle-road cooperative semantic communication scenario based on roadside computing network in this application;

[0054] Figure 2 This is a system flowchart of the vehicle-road cooperative communication system in the embodiments of this application;

[0055] Figure 3 This is a flowchart illustrating the vehicle-road cooperative communication method in the embodiments of this application;

[0056] Figure 4 This is a schematic diagram of the CNN-GRU prediction model in the embodiments of this application;

[0057] Figure 5 This is a schematic diagram of the vehicle trajectory prediction map in an embodiment of this application;

[0058] Figure 6 This is a schematic diagram illustrating the change in communication distance between the roadside unit (RSU) and the vehicle in an embodiment of this application.

[0059] Figure 7 This is a schematic diagram illustrating the linear relationship between semantic compression ratio and η(ο) in the embodiments of this application;

[0060] Figure 8This is a schematic diagram of the AOI variation curve of the vehicle received data in an embodiment of this application;

[0061] Figure 9 This is a schematic diagram of the information availability rate changing with the compression ratio in the embodiments of this application;

[0062] Figure 10 This is a flowchart illustrating the framework of the SSC system in this application embodiment;

[0063] Figure 11 This is a flowchart illustrating the ISSC encoder in an embodiment of this application;

[0064] Figure 12 This is a schematic diagram of two consecutive Swing transformer network blocks in an embodiment of this application;

[0065] Figure 13 This is a schematic diagram of window partitioning and shifted window partitioning in the embodiments of this application;

[0066] Figure 14 This is a schematic diagram of the ISSC decoder in an embodiment of this application;

[0067] Figure 15 This is a schematic block diagram of the vehicle-road cooperative communication device in the embodiments of this application;

[0068] Figure 16 This is a schematic diagram of the structure of the computer device in the embodiments of this application. Detailed Implementation

[0069] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0070] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0071] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0072] Those skilled in the art will understand that the terms "client," "terminal," and "terminal device" as used herein include both devices that receive wireless signals, devices that only possess wireless signal receiver capabilities without transmission capabilities, and devices with receiving and transmitting hardware, devices that have receiving and transmitting hardware capable of bidirectional communication over a bidirectional communication link. Such devices may include: cellular or other communication devices such as personal computers or tablets, having single-line displays, multi-line displays, or cellular or other communication devices without multi-line displays; PCS (Personal Communications Service) that can combine voice, data processing, fax, and / or data communication capabilities; PDA (Personal Digital Assistant) that may include a radio frequency receiver, pager, internet / intranet access, web browser, notepad, calendar, and / or GPS (Global Positioning System) receiver; and conventional laptops and / or handheld computers or other devices that have and / or include radio frequency receivers. As used herein, "client," "terminal," and "terminal device" can be portable, transportable, installed in a means of transportation (air, sea, and / or land), or suitable and / or configured to operate locally and / or in a distributed manner, operating in any other location on Earth and / or in space. "Client," "terminal," and "terminal device" as used herein can also be a communication terminal, an internet access terminal, or a music / video playback terminal, such as a PDA, a MID (Mobile Internet Device), and / or a mobile phone with music / video playback capabilities, or a smart TV, set-top box, etc.

[0073] The hardware referred to by the names "server," "client," and "service node" in this application is essentially an electronic device with the equivalent capabilities of a personal computer. It is a hardware device with the necessary components revealed by the von Neumann architecture, such as a central processing unit (including an arithmetic logic unit and a control unit), memory, input devices, and output devices. The computer program is stored in its memory, and the central processing unit loads the program stored in the secondary storage into the main memory to run it, execute the instructions in the program, and interact with the input and output devices to complete specific functions.

[0074] It should be noted that the concept of "server" used in this application can also be extended to the case of server clusters. Based on the network deployment principles understood by those skilled in the art, the servers should be logically divided. Physically, these servers can be independent of each other but accessible through interfaces, or they can be integrated into a single physical computer or a computer cluster. Those skilled in the art should understand this flexibility and should not use it to constrain the implementation of the network deployment method in this application.

[0075] Unless otherwise expressly specified, one or more technical features of this application may be deployed on a server and accessed by a client through remote invocation of the online service interface provided by the server, or they may be directly deployed and run on a client for access.

[0076] Unless otherwise specified, the neural network models referenced or potentially referenced in this application may be deployed on a remote server and invoked remotely on the client, or deployed on a client with the capability to invoke directly. In some embodiments, when running on the client, the corresponding intelligence may be acquired through transfer learning in order to reduce the requirements on the client's hardware resources and avoid excessive consumption of the client's hardware resources.

[0077] Unless otherwise specified, all data involved in this application may be stored remotely on a server or on a local terminal device, as long as it is suitable for use by the technical solution of this application.

[0078] Those skilled in the art will understand that although the various methods in this application are described based on the same concept and thus present commonality among them, they can be performed independently unless otherwise specified. Similarly, the various embodiments disclosed in this application are all based on the same inventive concept; therefore, concepts expressed in the same way, as well as concepts that are appropriately changed for convenience but are expressed differently, should be understood equivalently.

[0079] Unless otherwise expressly stated, the various embodiments disclosed in this application can be combined in a cross-cutting manner to flexibly construct new embodiments, as long as such combination does not depart from the inventive spirit of this application and can meet the needs of the prior art or solve a certain deficiency in the prior art. Those skilled in the art should be aware of such modifications.

[0080] In vehicle-road cooperative scenarios, vehicles need to collect various information on traffic segments in real time, such as traffic flow, road obstruction, pedestrian crossing, and traffic light information. After acquiring the data, they combine it with AIGC (Autonomous Data Collection) to provide various services. However, because vehicles need to collect a large amount of real-time data, the data volume, communication distance, and communication capabilities will affect transmission latency, potentially leading to data expiration. To address these issues, this application has implemented corresponding work on the roadside unit (RSU) to meet the needs of vehicles.

[0081] Vehicle-to-infrastructure (V2I) communication is an application based on advanced wireless communication and integrating next-generation Internet technologies, aiming to fully realize real-time information interaction between vehicles and between vehicles and the road network. By collecting and fusing dynamic traffic information in vehicle traffic scenarios, V2I enables vehicles to obtain various traffic conditions in real time, assisting in autonomous driving or other safety control operations, as well as road collaborative management, thereby achieving efficient collaboration between people, vehicles, and roads. Its main goals are to improve traffic safety, enhance traffic efficiency, and create a safe and efficient intelligent transportation system that serves society. Currently, a large amount of research on V2I applications has been carried out globally. V2I scenarios mainly include roadside units (RSUs) and on-board units (OBUs). Roadside units are used for V2I (Vehicle-to-Infrastructure) communication and are key devices for realizing intelligent road and vehicle-road collaboration. They are installed on the roadside and communicate and exchange data bidirectionally with nearby passing vehicles. In vehicle-road cooperative scenarios, the Onboard Unit (OBU) needs to acquire real-time road condition information, including scene and weather data. By deploying AIGC (Artificial Intelligence Generated Content) and other AI-generated content technologies on the OBU, the received semantic data can be linked with semantic graphs using AIGC multimodal technology to obtain realistic scene maps, thus assisting in autonomous driving. The Roadside Unit (RSU) can connect to existing roadside electronic devices, such as high-definition cameras, to acquire relevant data and upload it to the cloud or distribute it to vehicles—a crucial role of the RSU, especially on highways. Highway RSUs integrate perception units and connect to devices like high-definition cameras, enabling them to detect and identify abnormal road conditions—such as pedestrians or animals entering the road, vehicles stopping in the emergency lane, vehicles driving in the wrong direction, vehicle breakdowns, and traffic accidents. This collected image information can then be uploaded to the cloud in real time and simultaneously disseminated in advance to vehicles equipped with onboard terminals that are about to approach abnormal road sections. However, the large amount of image data interaction between the RSU and OBU is very time-consuming and puts a burden on the network. Therefore, semantic communication is introduced to extract semantic data from the raw data and express the desired meaning, filtering out irrelevant information from the original image to reduce information redundancy. After receiving semantic data from the RSU, the on-board unit provides various services through artificial intelligence content generation technologies such as AIGC.

[0082] Please see Figure 1 In a vehicle-road cooperative semantic communication application scenario based on roadside computing power network, where,

[0083] (1) The roadside unit (RSU) has information collection and sensing capabilities. It is connected to multiple high-definition cameras on the highway. The cameras collect information on traffic flow, road obstruction, pedestrian crossing, traffic light, etc. in real time.

[0084] (2) The Roadside Unit (RSU) has predictive capabilities. By introducing a lightweight CNN-GRU prediction model, the next-period position of the vehicle is predicted, further obtaining the vehicle's communication distance and channel resources. The optimal compression ratio is selected by combining data AOI and data fidelity to maximize the transmission success rate.

[0085] (3) The roadside unit (RSU) has image segmentation semantic compression capability. The roadside unit (RSU) is equipped with an image segmentation semantic communication system encoder. The image is semantically compressed according to the selected semantic compression ratio. Semantic compression reduces the amount of data, which can enable vehicles to receive data sent by the roadside unit (RSU) faster and more stably.

[0086] (4) The on-board unit (OBU) has the ability to decode and reconstruct image data. After the on-board unit (OBU) receives the data sent by the roadside unit (RSU), it performs semantic-level reconstruction of the data through the decoder of the image segmentation semantic communication system. The reconstructed image can be combined with AIGC to complete various services.

[0087] Please see Figure 2 The vehicle-road cooperative communication method of this application can be implemented based on a vehicle-road cooperative communication system, which can be divided into five modules, including: roadside unit layer, intelligent inference layer, semantic representation layer, semantic communication layer, and vehicle unit layer. The system processes data in a top-to-bottom and left-to-right order, and the functions of each layer of the system are as follows:

[0088] Roadside Unit Layer: The roadside unit layer consists of various roadside units (RSUs). Each RSU is connected to numerous electronic devices, such as high-definition cameras, responsible for collecting relevant data. They can sense and identify abnormal conditions on the road surface, such as pedestrians or animals crossing the road, vehicles parked in emergency lanes, vehicles driving in the wrong direction, vehicle breakdowns, and traffic accidents. This data is then uploaded to the cloud and distributed to vehicles to assist driving. Simultaneously, the RSUs can obtain historical driving trajectories and channel information from the cloud, serving as input for the intelligent inference layer.

[0089] Intelligent Inference Layer: The intelligent inference layer consists of the CNN-GRU prediction model. Based on the vehicle's historical driving trajectory, the GRU uses the prediction model to obtain the vehicle's position in the next period, and further obtains the vehicle's communication distance and channel resource status. Among them, the channel resources are related to the traffic flow at the vehicle's location.

[0090] Semantic representation layer: After obtaining the vehicle's communication distance and channel resource status, the maximum data transmission success rate and AOI are modeled respectively. Among them, this application analyzes the constraints of data AOI and data fidelity in the transmission success rate expression, which affect the data transmission success rate. Under the condition of fixed communication status, these constraints are only related to the semantic compression ratio. Therefore, the goal of this application is to select the optimal compression ratio to maximize the data transmission success rate.

[0091] Semantic Communication Layer: In the semantic communication layer, the roadside unit (RSU) is responsible for image segmentation and semantic compression of the raw data. It extracts semantic features from the image data through the ISSC encoder and generates a semantic map according to a certain compression ratio. Through image semantic segmentation, the main feature information of the image is preserved, which can reduce the amount of data. After completing the semantic compression, the roadside unit (RSU) sends the data to the vehicle unit. The vehicle unit decodes and reconstructs the received data and reconstructs the image through a generative network.

[0092] Onboard Unit Layer: The onboard unit layer consists of onboard units (OBUs). The onboard units (OBUs) in the onboard unit layer obtain semantically compressed data from roadside units (RSUs), reconstruct the image through the ISSC decoder, and obtain semantically restored image data. This data is then combined with AIGC to realize various services.

[0093] Based on the above exemplary scenarios, please refer to Figure 3 In one embodiment of the vehicle-road cooperative communication method of this application, the method includes:

[0094] Step S10: Acquire traffic scene image data, vehicle historical trajectory data, and vehicle driving data, wherein the vehicle driving data includes vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle.

[0095] The roadside unit (RSU) can acquire traffic scene image data, vehicle historical trajectory data, and vehicle driving data. The traffic scene image data includes one or more of the following: traffic flow image data, road barrier image data, pedestrian crossing image data, and traffic light image data. The vehicle driving data includes vehicle speed, acceleration, left lane distance, right lane distance, and vehicle driving angle. The vehicle historical trajectory data is multiple location information of the current vehicle in the previous period. The left lane distance refers to the distance between the left wheel of the vehicle and the left lane line of the driving lane, and the right lane distance refers to the distance between the right wheel of the vehicle and the right lane line of the driving lane.

[0096] Specifically, the Roadside Unit (RSU) has information collection and perception capabilities. It connects to multiple high-definition cameras on the highway. The cameras collect real-time information on traffic flow, road obstacles, pedestrian crossings, traffic lights, and vehicle driving data such as vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle. At the same time, the Roadside Unit (RSU) can obtain historical driving trajectories and channel information from the cloud as input to the intelligent inference layer.

[0097] Step S20: Using a preset vehicle position prediction model, determine the vehicle position coordinates in the next period based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle; and determine the communication distance and channel resources between the roadside unit and the vehicle-mounted unit based on the vehicle position coordinates.

[0098] After acquiring traffic scene image data, vehicle historical trajectory data, and vehicle driving data, the roadside unit (RSU) uses a preset vehicle position prediction model to determine the vehicle's position coordinates in the next period based on the vehicle historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle. The vehicle position prediction model is a CNN-GRU prediction model. Based on the vehicle position coordinates, the communication distance and channel resources between the roadside unit and the vehicle-mounted unit are determined.

[0099] Furthermore, the step of determining the vehicle's position coordinates in the next period using a preset vehicle position prediction model based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and the vehicle's driving angle includes:

[0100] Step S201: Input the vehicle historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle into a preset vehicle position prediction model, wherein the vehicle position prediction model is a CNN-GRU prediction model.

[0101] Step S203: In the CNN network structure, the CNN network structure includes a convolutional layer, a ReLU activation function layer, a max pooling layer, and a flattening layer. The convolutional layer uses a filter to generate a feature matrix based on the vehicle's historical trajectory data, the vehicle's speed, acceleration, distance to the left lane, distance to the right lane, and the vehicle's driving angle. The ReLU activation function layer is used to calculate the output of the feature matrix. The max pooling layer is used to reduce the size of the feature matrix. The flattening layer is used to flatten the output of the previous layer.

[0102] Step S205: Pass the features extracted by the CNN network structure into the GRU model to predict the vehicle's position coordinates in the next period.

[0103] Specifically, Convolutional Neural Networks (CNNs) are a type of feedforward neural network that incorporates convolutional computations. They are one of the representative algorithms of deep learning. The GRU model is a highly effective variant of the LSTM network; its structure is simpler, and its performance remains excellent. Therefore, GRU is a very popular network currently. As a variant of LSTM, the GRU model can also solve the long dependency problem in RNN networks. GRU has fewer parameters, resulting in faster training speeds, and it reduces the risk of overfitting. Furthermore, as a lightweight network model, it is more suitable for application in Roadside Units (RSUs) where computational resources are limited.

[0104] The CNN-GRU prediction model consists of two parts: the first part uses a convolutional neural network (CNN) to extract features from the time series, and the second part uses a GRU model to predict the extracted features.

[0105] Please see Figure 4 In the first part, the processed data is input into the CNN network structure, which consists of four layers: convolutional layer, ReLU activation function, max pooling layer and flattening layer. The convolutional layer uses a filter to generate a feature matrix, the ReLU activation function calculates the output of the feature matrix, max pooling is used to reduce the size of the matrix, and the flattening layer is used to flatten the output of the previous layer.

[0106] In the second part, the features extracted by the CNN are passed to the GRU model and prediction is made based on the GRU model. The structure consists of two layers, namely the GRU layer and the fully connected layer. The output of the fully connected layer is used as the prediction result to determine the vehicle's position coordinates in the next period.

[0107] Furthermore, the NGSIM data of US highway driving data collected by the US FHWA can be used as the dataset for training the prediction model. The historical trajectory data of the vehicle, as well as vehicle driving data such as vehicle speed v, acceleration a, left lane distance l, right lane distance r, and vehicle driving angle m, can be used as input. Based on the first 10 position information of the current vehicle in the previous period, the model predicts the next 10 position information of the current vehicle in the next period. The input of the CNN-GRU model is the historical trajectory data of the vehicle, as well as vehicle driving data such as vehicle speed v, acceleration a, left lane distance l, right lane distance r, and vehicle driving angle m. The output is the normalized position coordinate data of the vehicle in the next period.

[0108] In some embodiments, please refer to Figure 5 , Figure 5The diagram shows the motion trajectory of three vehicles with different trajectories after prediction by the prediction model. The experimental results verify the feasibility of the trajectory prediction scheme proposed in this application.

[0109] This application predicts the vehicle's location in the next time step using a CNN-GRU prediction model. Based on the vehicle's location information at the next moment, it obtains the communication distance between the roadside unit (RSU) and the vehicle, as well as the channel resource situation at the vehicle's location at the next moment. The channel resource situation is mainly determined by the traffic flow of the road segment. Higher traffic flow means more devices using base station channel resources for data transmission at the same time, resulting in reduced transmission bandwidth and power for individual vehicles, slower data processing speed, and impact on the data AOI received by the vehicle. The communication distance variation curve between the roadside unit (RSU) and the vehicle is shown in the figure. Figure 6 As shown.

[0110] Step S30: Determine the probability of successful data transmission and the probability of successful task execution under successful data transmission based on the communication distance and channel resources; construct an information availability probability model based on the probability of successful data transmission and the probability of successful task execution under successful data transmission.

[0111] After determining the communication distance and channel resources between the roadside unit and the vehicle-mounted unit based on the vehicle's location coordinates, the probability of successful data transmission and the probability of successful task execution under successful data transmission are determined based on the communication distance and channel resources. An information availability probability model is then constructed based on the probability of successful data transmission and the probability of successful task execution under successful data transmission.

[0112] The information can be used in a probability model as follows:

[0113] Φ=η(o)×P(a≤a0)

[0114] Where P(a≤a0) represents the probability of successful data transmission, a0 is the upper limit of the data AOI, η(o) represents the probability of successful task execution under successful transmission, o is the semantic compression ratio, and the data AOI is the sum of semantic compression processing delay, data transmission delay, and data propagation delay.

[0115] The steps to determine data transmission delay include:

[0116] Step S301: Determine the amount of traffic scene image data after semantic compression and the data transmission rate;

[0117] Step S303: Determine the transmission delay based on the ratio between the amount of traffic scene image data after semantic compression and the data transmission rate.

[0118] The steps to determine data propagation delay include:

[0119] Step S3001: Determine the communication distance and data propagation rate between the roadside unit and the vehicle-mounted unit;

[0120] Step S3003: Calculate and determine the data propagation delay based on the ratio between the communication distance between the roadside unit and the vehicle-mounted unit and the data propagation rate.

[0121] Specifically, image semantic compression ratio refers to the compression ratio between the size of the image data compressed by the encoder based on semantic information and the size of the original image data. In semantic communication models, using a higher compression ratio requires fewer channel resources but leads to semantic distortion, while using a lower compression ratio requires more channel resources. Therefore, excessively high AIGC (AI Compression Gear) may cause transmission failures. Thus, it is necessary to dynamically adjust the semantic compression ratio to maximize the information availability success rate. The information availability success rate (information availability probability model) is defined as follows:

[0122] Φ=η(o)×P(a≤a0) (1)

[0123] Where P(a≤a0) represents the probability of successful data transmission, a0 is the upper limit of the data AOI, η(o) represents the probability of successful task execution under successful data transmission, o is the semantic compression ratio, and the data AOI is the sum of semantic compression processing delay, data transmission delay, and data propagation delay.

[0124] For the probability η(ο) of successful task execution under successful data transmission, this application fits a linear relationship between semantic compression ratio and η(ο), such as... Figure 7 As shown, their linear relationship indicates that η(ο) is inversely proportional to the semantic compression ratio ο.

[0125] After the Roadside Unit (RSU) collects data, it performs semantic compression on the data before sending it to the vehicle. The entire process latency consists of three parts: semantic compression processing latency, transmission latency, and propagation latency. Regarding the semantic compression processing latency, since the image semantic compression method and compression coding algorithm used are consistent, the semantic compression processing latency can be considered to be within a stable interval, i.e., t... p =[t pmin , t pmax ].

[0126] The transmission delay is defined as follows:

[0127]

[0128] Where d = d0 × (1 - o) represents the amount of data after semantic compression, d0 represents the original length of the data, and v s This refers to the data transmission rate.

[0129] The propagation delay is defined as follows:

[0130]

[0131] Where l represents the communication distance, v r This represents the data propagation rate, according to Shannon's theorem.

[0132]

[0133] Where B is the bandwidth, P is the transmission power, h is the channel gain, and N0 is the noise power spectral density.

[0134] The AOI of the data is then defined as:

[0135] a = t p +t s +t r (5)

[0136] The probability of successful transmission is defined as follows:

[0137]

[0138] As can be seen from the above formula (6), the probability of successful transmission is mainly related to bandwidth, transmission power, communication distance, and semantic compression ratio. By obtaining the vehicle's bandwidth, transmission power, and communication distance through the vehicle location prediction model, when the bandwidth, transmission power, and communication distance are determined, the probability of successful transmission is only related to the semantic compression ratio and is directly proportional to it. Therefore, in order to maximize the information availability success rate of formula (1), it is necessary to dynamically select the optimal semantic compression ratio. The AOI of the vehicle's received data changes over time as shown in the figure. Figure 8 As shown; the curve of information availability changing with compression ratio is as follows. Figure 9 As shown.

[0139] Step S40: Determine the maximum data successful transmission probability, determine the optimal semantic compression ratio based on the information availability probability model and the maximum data successful transmission probability, and extract the semantic features of the traffic scene image data based on the ISSC encoder in the preset image segmentation semantic communication model according to the optimal semantic compression ratio to generate the semantic map corresponding to the traffic scene image data.

[0140] After constructing an information availability probability model based on the data success transmission probability and the probability of task success under successful data transmission, the maximum data success transmission probability is determined. Based on the information availability probability model and the maximum data success transmission probability, the optimal semantic compression ratio is determined. Based on the ISSC encoder in the preset image segmentation semantic communication model, the semantic features of the traffic scene image data are extracted according to the optimal semantic compression ratio to generate the semantic map corresponding to the traffic scene image data.

[0141] Furthermore, the step of the ISSC encoder in the preset image segmentation semantic communication model extracting semantic features of the traffic scene image data according to the optimal semantic compression ratio to generate a semantic map corresponding to the traffic scene image data includes:

[0142] Step S401: During the encoding process, the ISSC encoder extracts and aggregates semantic features from the traffic scene image data according to the optimal semantic compression ratio through a semantic feature extractor and a semantic feature aggregator. The semantic feature extractor is a Swin transformer network.

[0143] Step S403: The Swin transformer network consists of four stages, each stage having an even number of Swin transformer network blocks. The Swin transformer network blocks divide the input image into multiple independent small regions to determine window partitions, and calculate self-attention through a multi-head self-attention mechanism module based on window partitions.

[0144] Step S405: Each rule-based Swin transformer network block follows a multi-head self-attention mechanism module based on shifted window partitioning, which shifts the window partitioning. Pixels enable interaction between adjacent pixels in different window partitions to determine feature F. sc ;

[0145] Step S407, the feature F sc Feature fusion is performed after a fully connected layer. The compression ratio of the ISSC encoder is adjusted by the number of neurons in the fully connected layer to determine the output size of the semantic graph corresponding to the traffic scene image data.

[0146] Specifically, based on the information availability probability model and the maximum data successful transmission probability, the optimal semantic compression ratio is determined. Since the optimal semantic compression ratio is determined through the information availability probability model, image segmentation semantic compression can be performed. This application designs an Image Segmentation Semantic Communication System (ISSC), such as... Figure 10 As shown.

[0147] In this system, the transmitting end (roadside unit) transmits image semantic feature data to the receiving end (vehicle-mounted unit) via a wireless channel. The system consists of two parts:

[0148] The ISSC encoder at the transmitting end (roadside unit) extracts semantic features from the input image for transmission;

[0149] The ISSC decoder at the receiving end (vehicle unit) decodes the received semantic features and uses them to reconstruct image segmentation.

[0150] Please see Figure 11 During the encoding process, the ISSC encoder extracts and aggregates semantic features from the input image through a deep semantic feature extractor and a semantic feature aggregator. The input image is in RGB format, with each value ranging from [0, 255]. These values ​​need to be normalized to [0, 1] using a normalization layer to accelerate model convergence and reduce computational cost. Each image is then divided into 4×4 patches, as the dense prediction task at the receiving end (vehicle unit) benefits from using smaller patches. Considering that most previous research on image semantic communication systems is based on Convolutional Neural Networks (CNNs), whose semantic extraction capabilities are limited by the kernel size, the Swin transformer network possesses strong feature extraction capabilities and low computational complexity. Therefore, this application employs the Swin transformer network in the semantic feature extractor to extract deep semantic features from the image to generate a semantic map corresponding to the traffic scene image data.

[0151] Please see Figure 12 The Swin Transformer Network consists of four stages, each with an even number of Swin Transformer Network Blocks (STBs). The Swin Transformer Network Blocks (STBs) utilize a multi-head self-attention mechanism (MSA) to achieve a larger receptive field than CNNs. The architecture of the Swin Transformer Network Blocks (STBs) is as follows: Figure 12 As shown.

[0152] To reduce computational overhead, the Swing Transformer Network Block (STB) divides the input into many independent small regions, called window partitions, such as... Figure 13 As shown on the left, self-attention is calculated using a window-based multi-head self-attention mechanism (W-MSA) module. Furthermore, to enable interaction between adjacent pixels in different windows, each rule STB follows a shift-window-based multi-head self-attention mechanism (SW-MSA) module, which moves the window... Pixels, such as Figure 12 and Figure 13 As shown, the Swin Transformer Network Block (STB) also includes a Multilayer Perceptron (MLP) layer and two LayerNorm (LN) layers with residual connections.

[0153] Therefore, in the semantic feature extractor, through methods such as Figure 11 The three stages of deep feature extraction shown yield feature F. sc The image semantics of the features obtained in this process become increasingly richer. Finally, feature F scFeature fusion is performed after a fully connected layer. The compression ratio of the ISSC encoder is adjusted by the number of neurons in the fully connected layer, which determines the final output size of the image, and also the output size of the semantic graph corresponding to the traffic scene image data.

[0154] Step S50: Push the semantic map corresponding to the traffic scene image data to the vehicle unit. The ISSC decoder in the vehicle unit decodes the received semantic map corresponding to the traffic scene image data, reconstructs the image segmentation, and determines the data after image semantic segmentation to complete vehicle-road cooperative communication.

[0155] Based on the preset image segmentation semantic communication model, the ISSC encoder extracts the semantic features of the traffic scene image data according to the optimal semantic compression ratio to generate a semantic map corresponding to the traffic scene image data. Then, the semantic map corresponding to the traffic scene image data is pushed to the vehicle unit. The ISSC decoder in the vehicle unit decodes the received semantic map corresponding to the traffic scene image data, reconstructs the image segmentation, and determines the data after image semantic segmentation to complete vehicle-road cooperative communication.

[0156] The steps of the ISSC decoder in the vehicle unit decoding the semantic map corresponding to the received traffic scene image data and reconstructing image segmentation to determine the semantically segmented data include:

[0157] The ISSC decoder decodes the semantic graph corresponding to the received traffic scene image data, reconstructs the image segmentation, and determines the semantically segmented data. The ISSC decoder consists of a semantic feature decoder and a reconstructor.

[0158] The semantic feature decoder consists of a fully connected layer and two convolutional layers, and the reconstructor consists of two convolutional layers, three upsampling layers, a normalized activation layer, and an ArgMax layer.

[0159] Specifically, after the semantic map corresponding to the traffic scene image data is pushed to the vehicle-mounted unit, the ISSC decoder at the receiving end (vehicle-mounted unit) decodes the received semantic map corresponding to the traffic scene image data and reconstructs image segmentation. It consists of a semantic feature decoder and a reconstructor. The semantic feature decoder consists of a fully connected layer and two convolutional layers, as follows: Figure 14 As shown, this is to reduce the impact of noise. The reconstructor consists of two convolutional layers, three upsampling layers, one normalized activation (SoftMax) layer, and one ArgMax layer.

[0160] The goal of the reconstructor is to correctly classify image pixels. Therefore, this application uses the multi-class classification cross-entropy of each pixel as the loss function, expressed as:

[0161]

[0162] Where, p i is the probability that a pixel is classified into the i-th class, yi is the classification index with a value of 0 or 1, and N cls Let be the number of categories. For a batch of images, the loss function of the entire ISSC system is expressed as:

[0163]

[0164] Where B is the batch size of the total image data, H is the image height, and W is the image width. Given a loss function, the ISSC system is trained using an end-to-end method. Therefore, the semantic feature decoder can eliminate the impact on the channel when decoding the received semantic features.

[0165] In some embodiments, the On-Board Unit (OBU) obtains semantically compressed data from the Roadside Unit (RSU), reconstructs the image using an ISSC decoder to obtain semantically restored image data, and further combines AIGC to implement various services to complete vehicle-road cooperative communication between the Roadside Unit (RSU) and the On-Board Unit (OBU).

[0166] Compared to existing technologies, this application addresses the problems of poor data transmission stability, low transmission efficiency, and large data volume in vehicle-to-everything (V2X) communication, the lack of consideration for semantic compression ratio under data AOI constraints, and the weak feature extraction capability and high computational complexity of traditional image segmentation semantic communication models. This application offers the following advantages, including but not limited to:

[0167] Firstly, this application uses an image segmentation semantic communication model to process the pre-transmission data of vehicles, reducing the amount of data while ensuring data characteristics, thereby reducing the bandwidth resources consumed in the transmission process and maintaining a fast and reliable data sharing connection, so as to solve the problems of large data volume, communication cost and high latency in vehicle-road cooperation.

[0168] Secondly, the image segmentation semantic communication model of this application can guarantee strong feature extraction and low computational complexity, and achieve reliable fidelity before and after data transmission, so as to solve the problem of large data volume and high bandwidth consumption when transmitting data in vehicle-road cooperative systems.

[0169] Third, this application introduces a lightweight prediction model, CNN-GRU, into the roadside unit to predict the vehicle's position in the next period, further obtaining the vehicle's communication distance and channel resources. It also dynamically adjusts the semantic compression ratio by combining data AOI and data fidelity to maximize the success rate of information availability.

[0170] Furthermore, the image segmentation semantic communication system ISSC proposed in this application can solve the problems of large data volume and high bandwidth consumption when roadside units transmit data in vehicle-road cooperative systems, while ensuring strong feature extraction and low computational complexity, ensuring data fidelity, and reducing compression latency.

[0171] This application introduces a lightweight prediction model, CNN-GRU, into the roadside unit (RSU) to predict the next-period vehicle position, enabling more accurate tracking of real-time vehicle information. By obtaining vehicle communication status through location information, the application further models the data AOI, analyzes the relationship between data AOI, fidelity, and semantic compression ratio, and maximizes the success rate of information availability by dynamically adjusting the semantic compression ratio, which has practical application significance.

[0172] Please see Figure 15A vehicle-road cooperative communication device provided to meet one of the purposes of this application includes a data acquisition module 1100, a communication resource acquisition module 1200, an information availability model construction module 1300, a semantic graph generation module 1400, and a semantic graph push module 1500. The data acquisition module 1100 is configured to acquire traffic scene image data, vehicle historical trajectory data, and vehicle driving data, including vehicle speed, acceleration, left lane distance, right lane distance, and vehicle driving angle. The communication resource acquisition module 1200 is configured to use a preset vehicle position prediction model to determine the vehicle's position coordinates in the next period based on the vehicle historical trajectory data, vehicle speed, acceleration, left lane distance, right lane distance, and vehicle driving angle, and to determine the communication distance and channel resources between the roadside unit and the vehicle-mounted unit based on the vehicle position coordinates. The information availability model construction module 1300 is configured to determine the probability of successful data transmission and the probability of successful task execution under successful data transmission based on the communication distance and channel resources. An information availability probability model is constructed based on the probability of successful data transmission and the probability of successful task execution under successful data transmission. A semantic graph generation module 1400 is configured to determine the maximum probability of successful data transmission, determine the optimal semantic compression ratio based on the information availability probability model and the maximum probability of successful data transmission, and extract semantic features of the traffic scene image data from the ISSC encoder in a preset image segmentation semantic communication model according to the optimal semantic compression ratio to generate a semantic graph corresponding to the traffic scene image data. A semantic graph push module 1500 is configured to push the semantic graph corresponding to the traffic scene image data to the vehicle-mounted unit. The ISSC decoder in the vehicle-mounted unit decodes the received semantic graph corresponding to the traffic scene image data, reconstructs image segmentation, and determines the semantically segmented data to complete vehicle-road cooperative communication.

[0173] Based on any embodiment of this application, please refer to Figure 16 Another embodiment of this application also provides an electronic device, which can be implemented by a computer device, such as... Figure 16The diagram shows the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable storage medium stores an operating system, a database, and computer-readable instructions. The database may store control information sequences. When the computer-readable instructions are executed by the processor, the processor can implement a vehicle-to-infrastructure (V2I) cooperative communication method. The processor of the computer device provides computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can execute the V2I cooperative communication method of this application. The network interface of the computer device is used for communication with a terminal. Those skilled in the art will understand that… Figure 16 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0174] In this embodiment, the processor is used to execute... Figure 15 The system includes the specific functions of each module and its sub-modules, and the memory stores the program code and various data required to execute these modules or sub-modules. The network interface is used for data transmission between the user terminal and the server. In this embodiment, the memory stores the program code and data required to execute all modules / sub-modules in the vehicle-road cooperative communication device of this application, and the server can call the server's program code and data to execute the functions of all sub-modules.

[0175] This application also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the vehicle-road cooperative communication method described in any embodiment of this application.

[0176] This application also provides a computer program product, including a computer program / instructions that, when executed by one or more processors, implement the steps of the vehicle-road cooperative communication method described in any embodiment of this application.

[0177] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. This computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0178] The above description is only a partial embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

[0179] In summary, the Image Segmentation Semantic Communication System (ISSC) proposed in this application can address the challenges of large data transmission volumes and high bandwidth consumption in vehicle-road cooperative systems (V2L) while ensuring strong feature extraction and low computational complexity. It also guarantees data fidelity and reduces compression latency. By introducing a lightweight prediction model, CNN-GRU, into the Roadside Unit (RSU) to predict the vehicle's next-period position, it can more accurately track real-time vehicle information. The system obtains vehicle communication status through location information, further models the data AOI (Area of ​​Interest), analyzes the relationship between data AOI, fidelity, and semantic compression ratio, and maximizes information availability by dynamically adjusting the semantic compression ratio. This approach has practical application significance.

Claims

1. A vehicle-road cooperative communication method, characterized in that, include: Acquire traffic scene image data, vehicle historical trajectory data, and vehicle driving data, including vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle; A preset vehicle position prediction model is used to determine the vehicle's position coordinates in the next period based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle. Based on the vehicle position coordinates, the communication distance and channel resources between the roadside unit and the vehicle-mounted unit are determined. Based on the communication distance and channel resources, determine the probability of successful data transmission and the probability of successful task execution under successful data transmission, and construct an information availability probability model based on the probability of successful data transmission and the probability of successful task execution under successful data transmission. Determine the maximum probability of successful data transmission, determine the optimal semantic compression ratio based on the information availability probability model and the maximum probability of successful data transmission, and extract the semantic features of the traffic scene image data based on the optimal semantic compression ratio using the ISSC encoder in the preset image segmentation semantic communication model to generate a semantic map corresponding to the traffic scene image data. The semantic map corresponding to the traffic scene image data is pushed to the vehicle unit. The ISSC decoder in the vehicle unit decodes the received semantic map corresponding to the traffic scene image data, reconstructs the image segmentation, and determines the data after image semantic segmentation to complete vehicle-road cooperative communication.

2. The vehicle-road cooperative communication method according to claim 1, characterized in that, The step of determining the vehicle's position coordinates in the next period using a preset vehicle position prediction model based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and the vehicle's driving angle includes: The vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle are input into a preset vehicle position prediction model, which is a CNN-GRU prediction model. In the CNN network structure, the CNN network structure includes convolutional layers, ReLU activation function layers, max pooling layers, and flattening layers. The convolutional layer uses a filter to generate a feature matrix based on the vehicle's historical trajectory data, the vehicle's speed, acceleration, distance to the left lane, distance to the right lane, and the vehicle's driving angle. The ReLU activation function layer is used to calculate the output of the feature matrix. The max pooling layer is used to reduce the size of the feature matrix. The flattening layer is used to flatten the output of the previous layer. The features extracted by the CNN network structure are passed to the GRU model to predict the vehicle's position coordinates in the next period.

3. The vehicle-road cooperative communication method according to claim 1, characterized in that, The steps of determining the probability of successful data transmission and the probability of successful task execution under successful data transmission based on the communication distance and channel resources, and constructing an information availability probability model based on the probability of successful data transmission and the probability of successful task execution under successful data transmission, include: The information can be used in a probability model as follows: Φ=η(o)×P(a≤a0) Where P(a≤a0) represents the probability of successful data transmission, a0 is the upper limit of the data AOI, η(o) represents the probability of successful task execution under successful transmission, o is the semantic compression ratio, and the data AOI is the sum of semantic compression processing delay, data transmission delay, and data propagation delay.

4. The vehicle-road cooperative communication method according to claim 3, characterized in that, The steps to determine data transmission delay include: Determine the data volume and data transmission rate of traffic scene image data after semantic compression; The transmission delay is determined based on the ratio between the amount of semantically compressed traffic scene image data and the data transmission rate.

5. The vehicle-road cooperative communication method according to claim 3, characterized in that, The steps to determine data propagation delay include: Determine the communication distance and data propagation rate between the roadside unit and the vehicle-mounted unit; The data propagation delay is determined by calculating the ratio between the communication distance between the roadside unit and the vehicle-mounted unit and the data propagation rate.

6. The vehicle-road cooperative communication method according to claim 1, characterized in that, The steps of the ISSC encoder in the preset image segmentation semantic communication model to extract semantic features of the traffic scene image data according to the optimal semantic compression ratio to generate a semantic map corresponding to the traffic scene image data include: During the encoding process, the ISSC encoder extracts and aggregates semantic features from traffic scene image data according to the optimal semantic compression ratio through a semantic feature extractor and a semantic feature aggregator. The semantic feature extractor is a Swin transformer network. The Swin transformer network consists of four stages, each with an even number of Swin transformer network blocks. The Swin transformer network blocks divide the input image into multiple independent small regions to determine window partitions, and calculate self-attention through a multi-head self-attention mechanism module based on window partitions. Each rule-based Swin transformer network block follows a multi-head self-attention mechanism module based on shifted window partitions, which shifts the window partitions. Pixels enable interaction between adjacent pixels in different window partitions to determine feature F. sc ; The feature F sc Feature fusion is performed after a fully connected layer. The compression ratio of the ISSC encoder is adjusted by the number of neurons in the fully connected layer to determine the output size of the semantic graph corresponding to the traffic scene image data.

7. The vehicle-road cooperative communication method according to any one of claims 1 to 6, characterized in that, The ISSC decoder in the vehicle-mounted unit decodes the semantic map corresponding to the received traffic scene image data, and reconstructs the image segmentation to determine the semantically segmented data, including: The traffic scene image data includes one or more of the following: traffic flow image data, roadblock image data, pedestrian crossing image data, and traffic light image data. The ISSC decoder consists of a semantic feature decoder and a reconstructor; The semantic feature decoder consists of a fully connected layer and two convolutional layers, and the reconstructor consists of two convolutional layers, three upsampling layers, a normalized activation layer, and an ArgMax layer.

8. A vehicle-road cooperative communication device, characterized in that, include: The data acquisition module is configured to acquire traffic scene image data, vehicle historical trajectory data, and vehicle driving data. The vehicle driving data includes vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle. The communication resource acquisition module is configured to use a preset vehicle position prediction model to determine the vehicle position coordinates in the next period based on the vehicle's historical trajectory data, vehicle speed, acceleration, distance to the left lane, distance to the right lane, and vehicle driving angle, and to determine the communication distance and channel resources between the roadside unit and the vehicle-mounted unit based on the vehicle position coordinates. The information availability model construction module is configured to determine the probability of successful data transmission and the probability of successful task execution under successful data transmission based on the communication distance and channel resources, and to construct an information availability probability model based on the probability of successful data transmission and the probability of successful task execution under successful data transmission. The semantic graph generation module is configured to determine the maximum probability of successful data transmission, determine the optimal semantic compression ratio based on the information availability probability model and the maximum probability of successful data transmission, and extract the semantic features of the traffic scene image data based on the optimal semantic compression ratio using the ISSC encoder in the preset image segmentation semantic communication model to generate a semantic graph corresponding to the traffic scene image data. The semantic graph push module is configured to push the semantic graph corresponding to the traffic scene image data to the vehicle unit. The ISSC decoder in the vehicle unit decodes the received semantic graph corresponding to the traffic scene image data, reconstructs image segmentation to determine the data after image semantic segmentation, so as to complete vehicle-road cooperative communication.

9. An electronic device comprising a central processing unit and a memory, characterized in that, The central processing unit is used to invoke and run a computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores, in the form of computer-readable instructions, a computer program implemented according to any one of claims 1 to 7, which, when invoked by a computer, executes the steps included in the corresponding method.

Images