Vehicle track prediction method

Abstract

The invention discloses a vehicle trajectory prediction method, and the method comprises the steps: designing an intelligent collaborative architecture composed of a plurality of functional Agents, and enabling the intelligent collaborative architecture to be used for achieving the multi-level data analysis and semantic understanding in a vehicle trajectory prediction process; cooperative reasoning is realized among the Agents through task division and information interaction, so that multi-source dynamic information in a traffic scene is efficiently processed and comprehensively judged; meanwhile, a semantic communication mechanism is introduced, and in the V2V and V2I communication process, only original data is not transmitted any more, but high-level features and context information with semantic significance are transmitted, so that communication redundancy is remarkably reduced, and information transmission efficiency and semantic expression ability are improved; through semantic-level information interaction, the vehicle can more accurately understand the surrounding environment state and the potential traffic intention, so that the accuracy and robustness of trajectory prediction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation and vehicle networking communication, and in particular to a vehicle trajectory prediction method. BACKGROUND

[0002] With the continuous development of 5G and the upcoming 6G communication technology, the wireless communication capability, data transmission rate and network intelligence level have been significantly improved, which has promoted the emergence and innovation of various emerging application scenarios. Among them, the automatic driving technology is one of the typical high-bandwidth, high-reliability and low-latency application scenarios, and its development level directly reflects the overall capability of vehicle networking and intelligent transportation system. The core goal of automatic driving is to realize the running capability of vehicles in a complex and variable traffic environment with safe, efficient, intelligent and autonomous decision-making. Therefore, vehicles not only need to have their own perception, decision-making and control functions, but also need to interact with the surrounding environment in real time through multi-level communication networks. The vehicle networking (V2X) system is an important support for this goal, which includes vehicle-to-vehicle (V2V), vehicle-to-roadside unit (V2I), vehicle-to-pedestrian (V2P) and vehicle-to-network (V2N) communication modes, and realizes efficient use and intelligent scheduling of global traffic information through data sharing and collaborative decision-making.

[0003] However, vehicle trajectory prediction in complex traffic scenarios still faces multiple challenges. First, the traffic environment has high dynamicity and uncertainty, and vehicle behavior is affected by driving intention, road structure, signal control, surrounding vehicle state and other factors, making the trajectory prediction problem present obvious nonlinear characteristics. Second, although traditional deep learning methods can model the time sequence of vehicle motion trajectories, they mostly only focus on the historical data of the target vehicle itself, lacking the ability to deeply understand and model the semantic information of the surrounding environment. When the surrounding traffic state changes, the trajectory prediction model often cannot respond in time, resulting in a decrease in prediction accuracy and stability. In addition, the traditional communication system in vehicle networking mainly focuses on bit-level accuracy in information transmission, that is, how to accurately transmit data packets in a noisy channel, ignoring the "meaning" carried by the data. In a complex automatic driving environment, vehicles exchange not only data, but also context semantics and environmental understanding directly related to driving decision-making. If the communication system only targets bit accuracy and ignores the information value at the semantic level, it will waste communication resources, limit the understanding ability of intelligent agents to the environment, and affect the effectiveness of trajectory prediction.

[0004] Therefore, it is necessary to develop a vehicle trajectory prediction method to solve the above problems. SUMMARY

[0005] The purpose of the present application is to design a vehicle trajectory prediction method to solve the above problems.

[0006] The present application realizes the above-mentioned purposes through the following technical solutions:

[0007] A vehicle trajectory prediction method, comprising:

[0008] In V2I: performing first feature extraction on historical trajectory information of all vehicles within the communication range of the RSU; performing first semantic analysis on the first feature extraction result of the historical trajectory information of all vehicles within the communication range of the RSU; and predicting future trajectory information of the vehicle itself by using the historical trajectory information of the vehicle itself, the first feature extraction result and the first semantic analysis result;

[0009] In V2V: performing second feature extraction on historical trajectory information of the vehicle itself; performing second semantic analysis on the feature extraction result of the historical trajectory information of the vehicle itself; and predicting future trajectory information of the vehicle itself by using future trajectory information of other vehicles, the second feature extraction result and the second semantic analysis result.

[0010] The present application has the following advantages:

[0011] The LLM technology and semantic communication technology are used to predict the future trajectory of the vehicle, the environmental perception, logical reasoning and autonomous decision-making ability of the Agentic AI are comprehensively considered, an intelligent collaborative architecture composed of multiple functional Agents is designed, which is used for realizing multi-level data analysis and semantic understanding in the vehicle trajectory prediction process; the Agents are cooperatively reasoned through task division and information interaction, so that the multi-source dynamic information in the traffic scene is efficiently processed and comprehensively judged; at the same time, the semantic communication mechanism is introduced, in the communication process of V2V and V2I, not only the original data is transmitted, but also the high-level features and context information with semantic meaning are transmitted, so that the communication redundancy is significantly reduced, and the information transmission efficiency and semantic expression ability are improved; through the semantic level information interaction, the vehicle can more accurately understand the surrounding environment state and potential traffic intention, so as to improve the accuracy and robustness of the trajectory prediction;

[0012] By designing the feature extraction Agent, the semantic analysis Agent and the trajectory prediction Agent, high-precision trajectory prediction of the vehicle in a complex traffic environment is realized; the method comprehensively utilizes V2I and V2V data, combines the reasoning ability of the large language model, fully understands the automatic driving environment, and thus provides more comprehensive environment perception and behavior prediction support for the target vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0013] Fig. 1 The system scene model diagram of the present application;

[0014] Fig. 2 The system diagram of trajectory prediction in V2I of the present application;

[0015] Fig. 3 System diagram for trajectory prediction in V2V of the present application. DETAILED DESCRIPTION

[0016] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0017] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative work fall within the scope of protection of the present application.

[0018] It should be noted that: similar reference numerals and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0019] In the description of the present application, it should be understood that the terms "upper", "lower", "inner", "outer", "left", "right", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship commonly placed when the product of the present application is used, or the orientation or positional relationship commonly understood by those skilled in the art, and are only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the indicated device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0020] In addition, the terms "first", "second", etc. are only used for differentiation in description, and cannot be understood as indicating or implying relative importance.

[0021] In the description of the present application, it should also be noted that, unless otherwise explicitly specified and limited, the terms "provided", "connected" and the like should be understood broadly, for example, "connected" can be fixedly connected, or detachably connected, or integrally connected; can be mechanically connected, or electrically connected; can be directly connected, or indirectly connected through an intermediate medium, or the internal communication of two elements. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0022] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0023] As Figs. 1-3 shown, a vehicle trajectory prediction method comprises:

[0024] S1, feature extraction Agent.

[0025] In V2I, considering that the surrounding environment information will affect the accuracy of vehicle trajectory prediction, the application extracts the traffic environment information in the communication range of RSU, extracts the core feature information that can reflect the road state, traffic distribution and traffic dynamics, and through the extraction of these global environmental features, the system can fully utilize the change information of the surrounding vehicles and road conditions in the prediction stage, thereby enhancing the sensitivity and adaptability of trajectory prediction to the external environment.

[0026] In V2V, in order to fully understand the vehicle trajectory information, the application extracts the information of the target vehicle itself to assist in predicting the future trajectory of the vehicle itself; the feature extraction Agent first loads the historical trajectory information of the vehicle, and screens out key features therefrom, including the spatiotemporal dynamic features of the vehicle motion state, the statistical features representing the trajectory distribution characteristics, and the behavior intention features representing the vehicle in the road environment.

[0027] S2, semantic analysis Agent

[0028] In V2I, in order to fully understand the surrounding environment information, the application introduces a semantic analysis Agent. The semantic analysis Agent performs semantic layer analysis on the vehicle trajectory features in the communication range of RSU, and its core task is to understand the dynamic relationship between vehicles, the interaction intention and the semantic features of the overall traffic state from the global traffic perspective; in this way, the system can extract representative semantic information from group traffic behavior to provide semantic support for subsequent trajectory prediction. This process not only focuses on the motion change of a single vehicle, but also comprehensively considers the spatial correlation and behavior mode between vehicles, thereby realizing global semantic reasoning with more context understanding ability in the prediction stage.

[0029] In V2V, the semantic analysis Agent mainly faces the feature data of the target vehicle itself. By analyzing the historical trajectory features and adjacent vehicle behavior information, the semantic analysis Agent can understand the driving tendency, potential intention and behavior change law of the vehicle under different environmental conditions. Through this semantic layer behavior understanding, the system can more accurately grasp the potential decision trend of the target vehicle under the current traffic state, and provide fine-grained semantic input for trajectory prediction.

[0030] S3, semantic communication

[0031] In the communication between RSU and vehicle system, semantic transmission is adopted between RSU and vehicle. RSU extracts features of global traffic environment information by using feature extraction Agent, and performs upper and lower semantic reasoning and analysis on the feature extraction result by using semantic analysis Agent. The feature extraction result and the semantic analysis result are subjected to semantic encoding by a semantic encoder, and the semantic encoding result is subjected to channel encoding by a channel encoder, and is transmitted to the vehicle end through a wireless channel. After receiving the signal, the vehicle obtains the semantic encoding information through a channel decoder, and decodes and reconstructs the received decoded semantic encoding information by using a semantic decoder, and finally restores the global environment state information with clear semantic meaning, thereby providing auxiliary information for trajectory prediction.

[0032] In the communication between vehicles, semantic communication is adopted between vehicles, and the prediction result of the adjacent vehicle is transmitted to the target vehicle itself through semantic transmission. The target vehicle itself predicts the future trajectory information of the target vehicle itself by using the prediction result and the feature extraction result and the semantic analysis result of its own information. Specifically, after receiving the information transmitted by the adjacent vehicle, the target vehicle itself first decodes the information by using a channel decoder, and then decodes and reconstructs the channel decoded information by using a semantic decoder, and finally restores the trajectory information with clear semantic meaning. The target vehicle itself extracts features of its own information by using a feature extraction Agent, and performs upper and lower semantic reasoning and analysis on the feature extraction result by using a semantic analysis Agent. The target vehicle itself predicts the trajectory of the target vehicle itself by using the obtained feature extraction result, semantic analysis result and received trajectory information. After obtaining the prediction result, the target vehicle itself transmits the prediction result to the adjacent vehicle through semantic communication. The semantic communication system adopted in this part is consistent with V2I.

[0033] In order to reduce semantic distortion in the process of wireless transmission, the semantic encoding is performed by a semantic encoder , and the semantic encoding process can be represented as

[0034]

[0035] wherein, is the original vector to be transmitted, is the semantic encoding, is the bandwidth ratio of the channel, is the parameter of the semantic encoder .

[0036] When transmitted in a wireless fading channel, the complex vector will be affected by transmission loss including distortion and noise. The process can be represented as

[0037]

[0038] wherein, is the received complex vector, is the channel gain between the transmitter and the receiver, is the additive white Gaussian noise.

[0039] The semantic decoder receives the vector and reconstructs it, which can be represented as

[0040]

[0041] wherein, is the semantic decoder, is the parameter of the semantic decoder .

[0042] S3, Trajectory Prediction Agent

[0043] In V2I, in order to improve the prediction accuracy, the trajectory prediction agent uses the historical trajectory information of the vehicle itself, the feature extraction result and the global semantic analysis result to predict the future trajectory of the target vehicle. This process considers many factors such as traffic environment, surrounding vehicle behavior and road conditions, so that the prediction result can reflect the dynamic change trend of the vehicle in the overall traffic environment. By introducing environmental semantic information, the system can realize the transition from single vehicle behavior prediction to trajectory reasoning based on scene understanding, thereby improving the adaptability and accuracy of trajectory change in complex traffic state.

[0044] In V2V, the trajectory prediction agent analyzes the target vehicle's own historical trajectory and semantic information, and uses the predicted future trajectory of the adjacent vehicles within the communication range to realize collaborative prediction. This design fully considers the interaction between adjacent vehicles, so that the system can dynamically adjust the response strategy to the behavior of surrounding vehicles in the prediction process. Through this collaborative prediction mechanism, the system can more comprehensively understand the potential behavior relationship between vehicles, significantly improving the consistency and stability of trajectory prediction in complex traffic interaction scenarios.

[0045] This method introduces a multi-agent collaborative mechanism, designs a multi-functional agent system including feature extraction agent, semantic analysis agent and trajectory prediction agent, and realizes efficient trajectory prediction in complex traffic scenarios through task decomposition and collaborative optimization. At the same time, the invention combines semantic communication technology with multi-agent mechanism to realize semantic-level information transmission in the process of vehicle-to-vehicle and vehicle-to-road cooperative communication, thereby improving the information transmission efficiency and understanding accuracy. Through collaborative reasoning and semantic sharing between agents, vehicles can accurately obtain the trend of changes in the surrounding environment under low-latency communication conditions, and realize high-precision prediction of future trajectory.

[0046] The above merely is the preferred embodiment of the present application, it should be pointed out that, for ordinary skilled in the art, without departing from the technical principles of the present application, can also make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A vehicle trajectory prediction method, characterized in that, include: In V2I: First feature extraction is performed on the historical trajectory information of all vehicles within the RSU communication range; First semantic analysis is performed on the first feature extraction results of historical trajectory information of all vehicles within the RSU communication range; the vehicle's own historical trajectory information, first feature extraction results, and first semantic analysis results are used to predict the vehicle's own future trajectory information; In V2V: Second feature extraction is performed on the vehicle's own historical trajectory information; The feature extraction results of the vehicle's own historical trajectory information are subjected to second semantic analysis; the future trajectory information of other vehicles, the second feature extraction results, and the second semantic analysis results are used to predict the future trajectory information of the vehicle itself.

2. The vehicle trajectory prediction method according to claim 1, characterized in that, In V2I: the extracted features are core feature information that can reflect road conditions, traffic flow distribution and traffic dynamics; in V2V: the vehicle's own historical trajectory information is loaded, and key features are selected from it, including the spatiotemporal dynamic features of the vehicle's motion state, statistical features representing trajectory distribution characteristics, and feature information representing the vehicle's behavioral intentions in the road environment.

3. The vehicle trajectory prediction method according to claim 2, characterized in that, In V2I, a semantic analysis agent is introduced. The semantic analysis agent performs semantic layer analysis on the vehicle trajectory features within the RSU communication range. From a global traffic perspective, it understands the dynamic relationships, interaction intentions, and semantic features of the overall traffic state between vehicles. The system extracts representative semantic information from the group traffic behavior to provide semantic support for subsequent trajectory prediction. This process not only focuses on the motion changes of individual vehicles, but also comprehensively considers the spatial associations and behavioral patterns between vehicles. In V2V, the semantic analysis agent primarily focuses on the characteristic data of the target vehicle itself. By analyzing the target vehicle's own historical trajectory characteristics and the behavior information of surrounding vehicles, the semantic analysis agent understands the vehicle's driving tendencies, potential intentions, and behavioral change patterns under different environmental conditions.

4. The vehicle trajectory prediction method according to claim 3, characterized in that, In the communication between the RSU and the vehicle system, semantic transmission is used between the RSU and the vehicle. The semantic encoder performs semantic encoding on the feature extraction results and semantic analysis results, and the channel encoder performs channel encoding on the semantic encoding results. The signal is then transmitted to the vehicle via a wireless channel. After receiving the signal, the vehicle obtains the semantic encoding information through the channel decoder. The semantic decoder then performs semantic decoding and reconstruction on the received decoded semantic encoding information, and finally recovers the global environmental state information with clear semantic meaning, providing auxiliary information for trajectory prediction. In vehicle-to-vehicle communication, semantic communication is used between vehicles. Neighboring vehicles send their prediction results to the target vehicle via semantic transmission. The target vehicle itself uses the second feature extraction result and the second semantic analysis result to predict its own future trajectory information. Specifically, after receiving the information transmitted by the neighboring vehicle, the target vehicle first uses a channel decoder to perform channel decoding, and then uses a semantic decoder to perform semantic decoding and reconstruction of the channel decoded information, finally recovering trajectory information with clear semantic meaning. The target vehicle uses a feature extraction agent to extract features from its own information, and uses a semantic analysis agent to perform semantic reasoning and analysis on the feature extraction results. After predicting its own future trajectory information and obtaining the prediction result, the target vehicle sends the prediction result to other neighboring vehicles through semantic communication.

5. The vehicle trajectory prediction method according to claim 4, characterized in that, Semantic encoding is achieved through a semantic encoder. The semantic encoding process can be represented as follows: ； in, It is the original vector of the emission. It is semantic encoding. It is the bandwidth ratio of the channel. It is a semantic encoder Parameters; When transmitted over a wireless fading channel, complex vector The transmission loss can be caused by distortion and noise, and this process can be represented as: ； in, It is the received complex vector. The channel gain between the transmitter and receiver. It is additive white Gaussian noise; The semantic decoder receives the vector Then, it will be reconstructed, a process that can be represented as: ； in, For semantic decoders, For semantic decoders The parameters.

6. The vehicle trajectory prediction method according to claim 5, characterized in that, In V2I, the trajectory prediction agent uses the vehicle's own historical trajectory information, the results of the first feature extraction, and the results of the first semantic analysis to predict the vehicle's future trajectory information. This process comprehensively considers traffic environment, surrounding vehicle behavior, and road conditions, so that the prediction results can reflect the dynamic change trend of the vehicle in the overall traffic environment. In V2V, the trajectory prediction agent uses the future trajectory information of other vehicles, the results of second feature extraction, and the results of second semantic analysis to predict the future trajectory information of the vehicle itself. This process takes into account the interaction between neighboring vehicles, enabling the system to dynamically adjust its response strategy to the behavior of surrounding vehicles during the prediction process.

Images