Vehicle-to-everything (v2x) perception sharing message triggering method and system based on dynamic scene cognition

Abstract

The application provides a vehicle-end V2X perception sharing message triggering method and system based on dynamic scene cognition, and belongs to the technical field of intelligent networked vehicles and vehicle-road cloud integration. The application divides V2X messages into basic traffic messages and safety class perception sharing messages, and focuses on on-demand triggering of safety class messages. The vehicle-end fuses and collects data through multiple source sensors, adopts a state machine and a time sequence model to realize dynamic updating of scene cognition results, accurately identifies complex scenes such as occluded blind areas and sudden dangers, and predicts the change trend thereof; the sending end dynamically triggers perception information sharing based on the scene risk level, the receiving end initiates a request according to its own perception short board, and a communication load dynamic adaptation strategy is introduced. The application can significantly reduce the V2X channel load, expand the vehicle perception range, solve the problems of perception lag, invalid message redundancy and the like in the prior art, and realize the collaborative linkage of vehicle perception resources and intelligent traffic related subjects.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected vehicles and vehicle-road-cloud integration technology, and relates to a method and system for triggering V2X perception sharing messages based on dynamic scene cognition on the vehicle side. Background Technology

[0002] With the in-depth development of the vehicle-road-cloud integrated system, the penetration rate of V2X terminals continues to rise, L2+ level intelligent connected vehicles are gradually becoming widespread, the performance of vehicle-mounted sensing devices is constantly being upgraded, and the vehicle-mounted sensing capabilities are being enhanced day by day. As a result, the volume of V2X communication messages has exploded, bringing huge communication pressure to the V2X channel and becoming a key bottleneck restricting the safe passage of intelligent connected vehicles.

[0003] In existing technologies, V2X communication generally adopts a "periodic broadcast" mode, failing to accurately distinguish message types. Regardless of whether the driving scenario requires it, it continuously sends perception messages to the surrounding area, resulting in a large number of invalid messages occupying channel resources, triggering a "message storm," causing communication congestion, increased transmission latency, and even preventing the timely transmission of safety-related messages, thus creating potential driving safety hazards. Furthermore, existing V2X collaborative perception solutions mostly employ single, static scenario determination logic, lacking the ability to adapt to dynamic changes in driving scenarios. They cannot differentiate message sending or requests based on real-time changes in occlusion status, environmental interference, and hazard levels, further reducing the effectiveness of message processing.

[0004] Furthermore, in existing technologies, vehicles rely heavily on passively receiving perception messages from other nodes to address blind spots and beyond-line-of-sight areas. They cannot proactively and accurately request high-value perception information from surrounding vehicles, roadside facilities, and the cloud based on their own perception needs. This results in delayed blind spot perception and an inability to respond promptly to sudden dangers. At the same time, onboard perception resources are not fully utilized, and perceived safety hazard information cannot be effectively linked with relevant departments such as municipal authorities, traffic management agencies, and third-party map providers. This fails to fully leverage the collaborative advantages of the vehicle-road-cloud integrated system and cannot meet the advanced safety requirements of L2+ and above intelligent connected vehicles.

[0005] Existing technologies, such as patents CN119967365A and CN119450373A, disclose a scenario-based V2X message triggering method. However, its scenario determination is based solely on a single dimension of hazard level, failing to consider dynamic changes in environmental interference and communication load conditions. Furthermore, it only supports one-way proactive sharing, thus failing to address the core issues of message storms and insufficient perception accuracy. Therefore, there is an urgent need for a mechanism that can circumvent the limitations of existing technologies and achieve on-demand, accurate, and bidirectional triggering of safety-related V2X perception sharing messages based on dynamic scenario cognition. This mechanism should significantly reduce the volume of V2X messages and improve message processing efficiency and collaborative perception capabilities while ensuring driving safety. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a method and system for triggering V2X perception sharing messages based on dynamic scene cognition on the vehicle side. By constructing a dynamic scene cognition model, bidirectional intelligent triggering logic, hierarchical message scheduling, and perception closed-loop update, it realizes on-demand and accurate triggering of safety-related V2X perception sharing messages. Under the premise of ensuring driving safety, it significantly reduces the V2X channel load, expands the vehicle's perception range, makes full use of on-board perception resources, and improves the effectiveness of V2X message processing and the reliability of collaborative perception. This solves the technical defects in the prior art, such as frequent V2X message storms, static scene determination, insufficient accuracy of perception sharing, lagging processing of perception blind spots, and insufficient utilization of on-board perception resources.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for triggering V2X perception sharing messages based on dynamic scene cognition on the vehicle side, wherein the V2X perception sharing messages are safety-specific messages, distinct from basic passage messages, and are divided into proactive sharing and distribution messages triggered by the sender and targeted sharing request messages triggered by the receiver. The method specifically includes the following steps: S1. Vehicle-side multi-source data acquisition and dynamic fusion: Intelligent connected vehicles collect real-time traffic objects, road environment and their own operation data around the vehicle through on-board composite sensors. At the same time, they receive roadside perception data from roadside units through V2X communication and obtain global situational data from the cloud (including at least regional perception resource distribution information) through 5G / Uu interface. The multi-source data is time-stamped, coordinate system unified and target association calibrated. Combined with the dynamic changes of driving scene, Kalman filtering or sliding time window algorithm is used to realize real-time data fusion and status update, and output dynamic fused perception data. S2. Vehicle-side dynamic scene cognition: Based on the dynamic fusion perception data, a multi-dimensional scene judgment algorithm is used to output the dynamic judgment results of occlusion level, environmental interference level, dangerous object level and scene risk type in real time using a state machine model. The judgment results are updated in real time as the driving scene changes through a time series prediction method. The scene risk types include active safety hazard scenes and passive perception limited scenes. S3, Two-way intelligent trigger decision-making: S31. Active sharing and distribution triggering: When the scene recognition result determines that it is an active safety hazard scene and meets the preset hierarchical sharing conditions, the sending end is triggered to send a structured and simplified active sharing and distribution message to surrounding vehicles, roadside systems, and cloud management departments through one or more communication methods such as V2V, V2I, and V2N. The message sending priority is dynamically adjusted according to the hazard level. S32. Receiver-side directional sharing request trigger: When the scene perception result determines that it is a passive perception limited scene and meets the preset accurate request conditions, the receiver is triggered to send a directional sharing request message to vehicles, roadside units, and cloud platforms in the specified area through one or more communication methods such as V2V, V2I, and V2N, specifying the request scope and demand type, and prioritizing the node that covers the request area and has the best communication link quality as the request target. S4. Intelligent message scheduling and perception closed-loop update: The sending end performs hierarchical frequency control and lifecycle management on the proactively shared and distributed messages. After parsing and verifying the collaborative perception response data, the receiving end integrates it into the vehicle's dynamic perception model, updates the driving risk assessment and driving assistance strategies in real time, and dynamically adjusts the triggering logic based on the perception supplement results to form a closed-loop collaboration.

[0008] Furthermore, in step S1, the vehicle-mounted composite sensor includes a high-definition camera, a multi-mode millimeter-wave radar, a lidar, an IMU inertial measurement unit, and a GNSS high-precision positioning module. The traffic object data includes the type, location, relative speed, motion trajectory, and behavior prediction information of vehicles, pedestrians, and non-motorized vehicles. The road environment data includes light intensity, raindrop / fog concentration, road surface adhesion coefficient, road markings, and temporary road condition anomaly information.

[0009] Furthermore, in step S2, the specific process of dynamic scene cognition includes: S21. Occlusion Level Determination: Based on the vehicle's field of view, obstacle outline features, lane topology, and dynamic position information of neighboring vehicles, combined with the sensing range of onboard sensors, the type and level of blind spot occlusion, intersection occlusion, curve occlusion, and large vehicle occlusion are determined and divided into four levels: no occlusion, mild occlusion, moderate occlusion, and severe occlusion. The occlusion status changes are tracked in real time through a target tracking algorithm. S22. Environmental Interference Level Determination: Based on light intensity, raindrop / fog concentration, road surface adhesion coefficient, camera imaging quality, and electromagnetic interference intensity, it is divided into four levels: no interference, mild interference, moderate interference, and severe interference. The focus is on distinguishing between dynamic interference (sudden rainstorms, sudden changes in strong light, and temporary electromagnetic interference) and static interference (fixed obstructions and normal insufficient lighting). S23. Hazardous Object Level Determination: Based on target type, relative speed, distance, collision time (TTC), lane conflict probability, and behavior prediction results, a dynamic risk assessment algorithm classifies targets into four levels: general targets, targets of concern, hazardous targets, and emergency hazardous targets. Emergency hazardous targets are further assessed using real-time trajectory prediction parameters. The behavior prediction is based on an intent recognition model that considers the relationship between the target's historical trajectory and lane lines, or on a conflict detection algorithm that considers the target's motion state and road topology. S24. Scenario Risk Type Determination: When the level of dangerous object is greater than or equal to the target of concern, or when a sudden traffic incident is identified as posing a potential threat to surrounding vehicles, it is determined to be an active safety hazard scenario; when the level of occlusion is greater than or equal to moderate occlusion, or the level of environmental interference is greater than or equal to moderate interference leading to a decrease in perception accuracy, or when driving on road sections such as curves, tunnels, and intersections that are prone to generating blind spots in perception and cannot obtain complete perception information, it is determined to be a passive perception-restricted scenario. The two types of scenarios can coexist and can be switched in real time through a state machine.

[0010] Furthermore, in step S31, the hierarchical sharing conditions are: the level of dangerous object ≥ dangerous target; or, the occlusion level ≥ moderate occlusion and the environmental interference level ≥ moderate interference; or, an accident, construction, road obstacles, road collapse, pedestrian / animal crossing, or other sudden events are identified as posing a potential threat to vehicles approaching from behind; the active sharing and distribution message is structured and simplified data, which at least includes target ID, real-time location, speed, heading, type, danger level, confidence level, event type, occurrence location, and risk duration information, with different event types supplemented with exclusive feature parameters.

[0011] Furthermore, in step S32, the precise request conditions are: occlusion level ≥ moderate occlusion and the sensor itself cannot fill the blind spot; or, the perception confidence of a specific target is lower than a preset threshold and cannot be calibrated by its own data; or, the environmental interference level ≥ severe interference causing perception failure; or, driving to road sections such as curves, tunnels, and intersections that are prone to perception blind spots and needing to obtain beyond-line-of-sight information; the directional sharing request message carries at least the coordinates of the request area, request type, priority, effective time limit, and a description of its own perception shortcomings. The request type includes traffic object recognition, event detection, road environment perception, and beyond-line-of-sight road condition query.

[0012] Furthermore, in step S4, the hierarchical frequency control and lifecycle management specifically involves dividing time windows according to the level of the hazardous object: the time window for emergency hazardous targets is 100ms-300ms, the time window for hazardous targets is 300ms-500ms, and the time window for targets of concern is 500ms-1000ms. The same target or event is sent only once within the corresponding time window, and expired messages are automatically discarded to avoid message redundancy caused by repeated sending.

[0013] Furthermore, in step S5, the sending of the proactive sharing and distribution message and the targeted sharing request message are dynamically adapted in conjunction with the current V2X communication load, which is measured by the channel busy rate (CBR). When the CBR is lower than a preset threshold, all types of messages that meet the conditions are sent normally. When the CBR is higher than the preset threshold, messages related to urgent and dangerous targets are sent first, and the sending of non-urgent messages is temporarily suspended. At the same time, the messages are simplified a second time. The second simplification includes removing historical trajectory points, removing low-confidence environmental interference information, or aggregating multiple related targets into a group of information for sending.

[0014] This invention also provides a vehicle-side V2X perception sharing message triggering system based on dynamic scene cognition, used to implement the above method. The system includes: Multi-source dynamic fusion module: Composed of vehicle-mounted composite sensor interface, roadside data receiving interface, cloud data receiving interface and dynamic fusion unit, it is used to collect vehicle perception data, receive roadside and cloud data, complete spatiotemporal alignment, target association calibration and dynamic fusion of multi-source data, output dynamic fused perception data, and support sensor fault adaptive switching. Dynamic Scene Recognition Module: Connected to the multi-source dynamic fusion module, it has built-in occlusion level determination unit, environmental interference level determination unit, dangerous object level determination unit and scene risk type determination unit. It adopts a multi-dimensional dynamic judgment algorithm and state machine model to output scene recognition results in real time and update them synchronously. It also has a built-in scene prediction unit. Based on historical perception data and real-time dynamic fusion data, it uses time series prediction method to predict the trend of scene risk changes and trigger message sending or request preparation in advance to shorten the triggering delay. The bidirectional intelligent triggering module, connected to the dynamic scene cognition module, includes a sending-end hierarchical triggering unit, a receiving-end directional triggering unit, and a load dynamic adaptation unit. The sending-end hierarchical triggering unit triggers message sending according to the danger level based on the scene cognition results and hierarchical sharing conditions. The receiving-end directional triggering unit initiates directional requests based on the scene cognition results and precise request conditions, and optimizes the request object based on the regional perception resource distribution information distributed from the cloud. The load dynamic adaptation unit monitors the communication load (channel busy rate CBR) in real time and dynamically adjusts the message sending priority, sending timing, and message simplification level according to the load status. V2X intelligent communication module: connected to the bidirectional intelligent triggering module, supporting multiple communication modes such as V2V, V2I, and V2N, used to send active sharing and distribution messages and targeted sharing request messages, receive collaborative sensing response data, complete message protocol parsing, link monitoring, anomaly handling, and secondary message simplification to ensure transmission stability and real-time performance; Message scheduling and perception closed-loop module: connected to the V2X intelligent communication module, including a hierarchical frequency control unit, a lifecycle management unit, and a dynamic perception update unit. The hierarchical frequency control unit and the lifecycle management unit realize hierarchical message scheduling and expiration handling. The dynamic perception update unit integrates the received response data into the vehicle's dynamic perception model, updates the risk assessment and driving assistance strategies, and feeds back the perception update results to the dynamic scene cognition module and the bidirectional intelligent triggering module to adjust the scene cognition accuracy and triggering logic, forming a collaborative closed loop.

[0015] Furthermore, the load dynamic adaptation unit presets a low threshold and a high threshold for CBR. When CBR is lower than the low threshold, it is determined to be low load and normal transmission is performed. When CBR is higher than the high threshold, it is determined to be high load, triggering secondary simplification and suspending non-urgent messages. When CBR is between the two, it is determined to be medium load, and only low-priority messages are slightly simplified.

[0016] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the above-described vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition is implemented.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1) Precisely suppress V2X message storms and improve communication stability: By dividing V2X messages into basic communication messages and security-aware shared messages, and adopting a dynamic scene recognition triggering mechanism only for security messages, combined with hierarchical frequency control, message hierarchical simplification and communication load dynamic adaptation strategies (based on CBR threshold control), the amount of V2X messages can be reduced by 40%-75%, significantly reducing the V2X channel load, fundamentally suppressing message storms, and ensuring that security messages are transmitted in a priority and timely manner.

[0018] 2) Dynamic scene cognition to improve trigger accuracy: Construct a multi-dimensional dynamic scene cognition model, introduce environmental interference level judgment (distinguishing between dynamic / static interference) and dangerous object behavior prediction (based on intent recognition or conflict detection), and realize real-time updating and trend prediction of scene judgment results through state machine and time series prediction, avoiding problems such as untimely triggering and blind requests caused by static scene judgment in existing technologies, and ensuring the accuracy and timeliness of message triggering.

[0019] 3) Two-way closed-loop triggering to expand the perception range: Construct a two-way triggering logic of hierarchical sharing at the sending end and directional requests at the receiving end. Combined with the distribution of cloud perception resources, it realizes accurate requests, breaks the limitations of single-vehicle perception, effectively solves problems such as blind spots and insufficient beyond-line-of-sight perception, improves the vehicle's perception ability of unknown areas, reduces collision risks, and adapts to the high-level safe passage requirements of L2+ and above intelligent connected vehicles.

[0020] 4) Efficient utilization of vehicle-mounted sensing resources to enhance collaborative value: Accurately share safety hazard information perceived by vehicle-mounted sensors with other vehicles, intelligent transportation systems, and relevant departments such as municipalities, traffic management, and third-party maps. At the same time, combine cloud-based sensing resources to optimize request efficiency, realize the collaborative utilization of vehicle-mounted sensing resources and intelligent transportation systems, and improve the overall collaborative sensing capabilities of the vehicle-road-cloud integrated system.

[0021] 5) Strong compatibility and easy to scale up: Based on the existing V2X communication protocol stack and vehicle sensor hardware, no large-scale hardware modification is required. It supports adaptive switching for sensor failures and can be directly adapted to existing L2+ level intelligent connected vehicles, roadside systems and cloud platforms, reducing deployment costs.

[0022] 6) Closed-loop collaborative optimization to improve reliability: Construct a complete collaborative closed loop of "perception-cognition-triggering-sharing-request-update-feedback", shorten trigger latency through scenario prediction, and adjust and optimize trigger logic and perception accuracy through feedback to improve the real-time performance, accuracy and reliability of V2X perception sharing.

[0023] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of the overall process of the V2X perception sharing message triggering mechanism based on dynamic scene cognition on the vehicle side of the present invention. Figure 2 This is a schematic diagram of the module composition of the V2X perception sharing message triggering system based on dynamic scene cognition on the vehicle side of the present invention. Figure 3 This is a logical diagram illustrating the dynamic scene recognition and multi-dimensional judgment of the present invention; Figure 4 This is a flowchart of the bidirectional intelligent triggering decision-making and message hierarchical scheduling of the present invention; Figure 5 This is a schematic diagram illustrating the interaction between proactive sharing and distribution, targeted sharing requests, and perception closed-loop updates in an embodiment of the present invention. Detailed Implementation

[0025] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0026] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0027] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0028] See Figure 1 — Figure 5 This invention provides a method and system for triggering V2X perception sharing messages based on dynamic scene cognition on the vehicle side. The V2X perception sharing messages are safety-specific messages, distinct from basic passage messages, and are divided into proactive sharing and distribution messages triggered by the sender and targeted sharing request messages triggered by the receiver. The triggering mechanism specifically includes the following steps: Step S1: Vehicle-side multi-source data acquisition and dynamic fusion Intelligent connected vehicles use onboard composite sensors (including high-definition cameras, multi-mode millimeter-wave radar, lidar, IMU inertial measurement units, and GNSS high-precision positioning modules) to collect real-time data on the type, location, relative speed, trajectory, and behavior prediction of surrounding traffic objects (vehicles, pedestrians, and non-motorized vehicles), road environment data (light intensity, rain / fog concentration, road surface adhesion coefficient, road signs, and temporary road condition anomalies), and the vehicle's own operational data (speed, heading, braking status, and steering intention). Simultaneously, they receive data from roadside units (RSUs) via a V2X communication module. The roadside sensing data sent by the system is processed via the 5G / Uu interface to obtain global situational data (global map, traffic congestion information, abnormal event warnings, and regional sensing resource distribution) from the cloud platform. This multi-source data undergoes unified timestamp and coordinate system processing to achieve spatiotemporal alignment. A target association calibration algorithm is used to fuse the multi-source data. Furthermore, considering the dynamic changes in the driving scenario, Kalman filtering or sliding time window algorithms are employed to update the fused data in real time. The output is dynamic fused sensing data that comprehensively and in real-time reflects the current driving scenario, providing accurate data support for subsequent dynamic scenario cognition. This step introduces a dynamic fusion algorithm to enable real-time updates of the fused data as the scenario changes and adds cloud-based regional sensing resource distribution data to provide a basis for targeted requests from the receiving end.

[0029] Step S2: Vehicle-side dynamic scene recognition Based on the dynamically fused sensing data obtained in step S1, a multi-dimensional dynamic judgment algorithm is used, employing a state machine model to output four core dynamic judgment results in real time, specifically including: Occlusion Level Determination: Based on the vehicle's field of view, the outline features of obstacles (large vehicles, buildings, guardrails, trees), lane topology, and the dynamic position information of neighboring vehicles, combined with the sensing range of the onboard sensors, the occlusion type (blind spot occlusion, intersection occlusion, curve occlusion, large vehicle occlusion) and occlusion level are determined, categorized into four levels: no occlusion, slight occlusion, moderate occlusion, and severe occlusion. A target tracking algorithm (such as Kalman filtering) is used to track the dynamic changes in the occlusion state in real time (e.g., changes in the occlusion range due to the movement of large vehicles) to ensure the real-time accuracy of the determination results. Slight occlusion refers to partial occlusion of non-critical areas, which does not affect core perception; moderate occlusion refers to partial occlusion of critical areas, resulting in decreased core perception accuracy; severe occlusion refers to complete occlusion of critical areas, rendering core perception ineffective.

[0030] Environmental interference level assessment: Based on light intensity, raindrop / fog concentration, road surface adhesion coefficient, camera imaging quality, and electromagnetic interference intensity, environmental interference is divided into four levels: no interference, mild interference, moderate interference, and severe interference. The assessment focuses on distinguishing between dynamic interference (sudden heavy rain, sudden changes in strong light, temporary electromagnetic interference) and static interference (fixed obstructions, chronically insufficient lighting). Mild interference indicates a slight decrease in perception accuracy, not affecting core judgment; moderate interference indicates a significant decrease in perception accuracy, with some targets being abnormally identified; severe interference indicates a significant failure of perception capability, making effective target identification impossible.

[0031] Hazard level determination: Based on the type of traffic object (pedestrian, non-motorized vehicle, motorized vehicle), relative speed, distance to the vehicle, time to collision (TTC), lane conflict probability, and behavior prediction results, a dynamic risk assessment algorithm categorizes the hazard level of targets into four levels: general targets, targets of concern, hazardous targets, and emergency hazardous targets. Behavior prediction can be achieved through the following methods: intent recognition models (such as LSTM networks) based on the relationship between the target's historical trajectory and lane lines; conflict detection algorithms based on the target's motion state (speed, acceleration, rate of change of heading angle) and road topology; or direct judgment based on vehicle steering / braking signals carried in V2X messages. Emergency hazardous targets additionally incorporate real-time trajectory prediction parameters to anticipate hazard changes in advance.

[0032] Scenario Risk Type Determination: Based on the above three dynamic determination results, driving scenarios are divided into active safety hazard scenarios and passive perception-limited scenarios. These two types of scenarios can coexist and switch in real time through a state machine. When the level of the dangerous object is greater than or equal to the target of interest, or when the vehicle detects sudden traffic events such as accidents, construction, road obstacles (falling rocks, potholes, protrusions, spilled objects), road collapses, or pedestrians or animals crossing the road that pose a potential threat to surrounding vehicles (especially vehicles approaching from behind), it is determined to be an active safety hazard scenario. When the level of occlusion is greater than or equal to moderate occlusion, or the level of environmental interference is greater than or equal to moderate interference leading to a decrease in perception accuracy, or when the vehicle is driving on road sections that are prone to perception blind spots such as curves, tunnels, and intersections and cannot obtain complete perception information, it is determined to be a passive perception-limited scenario.

[0033] Step S3: Two-way intelligent trigger decision Based on the dynamic scenario awareness results from step S2, and combined with the current V2X communication load status, a bidirectional intelligent triggering decision is executed between the sending and receiving ends to achieve hierarchical and targeted message triggering, specifically including: S31. Active Sharing and Distribution Trigger: When the scene recognition result determines that it is an active safety hazard scene and meets the preset hierarchical sharing conditions, the active sharing and distribution logic of the sending end is triggered. The hierarchical sharing conditions are: dangerous object level ≥ dangerous target; or, occlusion level ≥ moderate occlusion and environmental interference level ≥ moderate interference; or, the vehicle identifies sudden events such as accidents, construction, road obstacles, road collapses, pedestrians or animals crossing the road that pose a potential threat to vehicles behind. After triggering, the sending end sends structured active sharing and distribution messages to surrounding vehicles, roadside systems, and cloud management departments (traffic management, municipal) through one or more of the following communication methods: V2V (vehicle-to-vehicle), V2I (vehicle-to-infrastructure), and V2N (vehicle-to-cloud). The proactively shared and distributed messages adopt a hierarchical and simplified format, adjusting message complexity according to the hazard level. They contain only core key parameters, including at least the target ID, real-time location, speed, heading, type, hazard level, confidence level, and information such as event type (e.g., accident, construction, road obstacle), location of occurrence, and duration of risk. For different emergency scenarios, specific feature parameters are added: road obstacles require markings of size and shape; road collapses require markings of collapse range and estimated depth; and pedestrian / animal crossings require markings of crossing trajectory, movement speed, and behavior prediction results. This avoids redundant data transmission. Simultaneously, message sending priority is dynamically adjusted according to the hazard level (emergency hazard targets have the highest priority).

[0034] S32. Receiver-side directional sharing request trigger: When the scene perception result determines that it is a passively limited perception scene, and the preset accurate request conditions are met, the receiver's directional sharing request logic is triggered. The accurate request conditions are: occlusion level ≥ moderate occlusion and the vehicle cannot fill the blind spot with its own sensors; or, the perception confidence of a specific target is lower than a preset threshold and cannot be calibrated with its own data; or, the environmental interference level ≥ severe interference causing the vehicle's perception capability to fail; or, driving on road sections such as curves, tunnels, and intersections that are prone to perception blind spots and needing to obtain beyond-line-of-sight information. After triggering, the receiver sends a directional sharing request message to vehicles, roadside units, and cloud platforms in the designated area through one or more communication methods such as V2V, V2I, and V2N. The message carries at least the coordinates of the requested area (such as blind spot, curve range), request type (traffic object recognition, event detection, road environment perception, beyond-line-of-sight road condition query), priority (urgent, normal), effective time limit, and a description of its own perception shortcomings to ensure the accuracy and timeliness of the request. This vehicle combines cloud-based regional perception resource distribution data to prioritize nodes (such as specific RSUs or neighboring vehicles) that cover the requested area and have the best communication link quality as the request target, thereby improving response efficiency.

[0035] Furthermore, when triggering message transmission, dynamic adaptation is performed based on the current V2X communication load: the communication load is measured by the Channel Busy Rate (CBR). A low CBR threshold (e.g., 30%) and a high CBR threshold (e.g., 60%) are preset. When the CBR is below the low threshold, all types of messages that meet the conditions are sent normally; when the CBR is above the high threshold, messages related to urgent and dangerous targets are prioritized for transmission, while the transmission of non-urgent messages is temporarily suspended. Simultaneously, messages are further simplified, retaining only core risk parameters (such as removing historical trajectory points, removing low-confidence environmental interference information, or aggregating multiple related targets into a group for transmission), further reducing the channel load; when the CBR is between the two, only low-priority messages are lightly simplified.

[0036] Step S4: Intelligent message scheduling and perception closed-loop update The sending end performs hierarchical frequency control and lifecycle management for proactively shared and distributed messages. Specifically, different time windows are divided according to the level of danger of the object: the time window for emergency dangerous targets is 100ms-300ms, the time window for dangerous targets is 300ms-500ms, and the time window for targets of interest is 500ms-1000ms. The same target or event is sent only once within the corresponding time window, and only the changed parameters of the target or event (such as changes in location, speed, and behavior status) are updated. Expired messages are automatically discarded to avoid message redundancy caused by repeated sending.

[0037] The receiving end receives collaborative perception response data returned from surrounding vehicles, roadside, and cloud via the V2X intelligent communication module. It analyzes and verifies the response data, removes invalid and abnormal data, and integrates it into the vehicle's dynamic perception model to supplement perception information in blind spots and beyond-line-of-sight areas. It updates driving risk assessment results in real time, optimizes driving assistance strategies (such as warning, deceleration, avoidance, and route adjustment), and improves vehicle driving safety.

[0038] Meanwhile, after receiving proactive sharing and distribution messages or collaborative perception response data, the receiving end determines whether it needs to initiate further targeted sharing requests or adjust request parameters based on the dynamic changes in its own scene cognition results; the sending end dynamically adjusts the message sending frequency and content based on the response feedback from the receiving end, forming a complete collaborative closed loop of "perception-cognition-triggering-sharing-request-update-feedback".

[0039] Example 1: Dynamic scenario of blind spots caused by large vehicles at urban intersections This embodiment corresponds to a scenario at an urban road intersection where a vehicle encounters a blind spot blocked by a large vehicle when turning right, and the blind spot dynamically changes as the large vehicle moves.

[0040] Multi-source data acquisition and dynamic fusion: This vehicle (L2+ level intelligent connected vehicle) is preparing to turn right at an urban intersection. It uses onboard composite sensors to collect data on the position and heading changes of a large truck on the right, its own data, lane markings at the intersection, and the location of pedestrian crossings. It receives dynamic data on pedestrian and non-motorized vehicle flow from the intersection's RSU via V2X, and acquires traffic signal status and regional perception resource distribution data from the cloud via 5G / Uu. The multi-source data undergoes spatiotemporal alignment and target association calibration. Kalman filtering is used to update the fused data in real time, resulting in dynamically fused perception data that clearly identifies the truck's obstruction range and blind spots.

[0041] Dynamic Scene Recognition: Based on dynamic fusion perception data, the occlusion level is determined in real time as severe occlusion (a large truck completely blocks the right-side view, and the occlusion range changes continuously as the truck moves), the environmental interference level is no interference, the dangerous object level is tentatively set as a target of interest (the target in the blind spot cannot be clearly perceived, but it is predicted that there may be pedestrians crossing the blind spot), and the scene risk type is simultaneously determined as a passive perception-limited scene and an active safety hazard scene. The scene recognition results are updated by tracking the changes in the occlusion range through a state machine.

[0042] Two-way intelligent trigger decision: (1) Receiver-side directional sharing request trigger: Due to the limitations of passive perception and severe occlusion, blind spots cannot be filled, and the conditions for accurate request are met, the receiver-side directional triggering unit triggers a directional sharing request. This vehicle combines the cloud-based regional perception resource distribution data and sends a directional sharing request message to the intersection RSU via V2I and to neighboring vehicles with perception capabilities upstream of the intersection via V2V. The message carries the request area (the right-side fan-shaped area occluded by the large vehicle, with coordinates updated in real time as the truck moves), request type (pedestrian and non-motorized vehicle recognition and behavior prediction), priority (high), effective time limit (2 seconds), and a description of its own perception shortcomings.

[0043] (2) Active sharing and distribution triggering at the sending end: When the vehicle receives blind spot perception data returned by the RSU (identifying one pedestrian and one electric vehicle in the blind spot, and predicting that the pedestrian intends to cross), the dynamic scene cognition module immediately updates the danger level of the object to an emergency danger target, meeting the conditions for hierarchical sharing. The hierarchical triggering unit at the sending end triggers active sharing and distribution, setting the message priority to the highest and the time window to 300ms. The vehicle sends active sharing and distribution messages to surrounding vehicles via V2V and to the roadside system via V2I, including structured and simplified data such as target ID, real-time location, speed, type, danger level, confidence level, and behavior prediction results. The current V2X communication load is low (CBR<30%), so normal transmission is possible without secondary simplification.

[0044] Intelligent message scheduling and perception closed-loop update: The sending end incrementally updates the actively shared and distributed messages according to a 300ms time window; after receiving the shared message, the receiving vehicle integrates it into its own dynamic perception model and issues an emergency warning for blind spots; after integrating the blind spot data returned by the RSU, the vehicle updates its own dynamic perception model, triggers active deceleration and steering adjustment, and feeds back the avoidance action to the dynamic scene cognition module to adjust the scene cognition results and postpone subsequent requests.

[0045] Example 2: Dynamic Interference Scenario of High-Speed ​​Fog This embodiment corresponds to a highway fog scenario, where dynamic changes in fog concentration lead to changes in the level of environmental interference, and there are also urgent and dangerous targets.

[0046] Multi-source data acquisition and dynamic fusion: When the vehicle is traveling at high speed and suddenly encounters dense fog, it collects data on the dynamic changes in fog concentration and the position and speed of the vehicle in front through onboard sensors; it receives data from neighboring vehicles ahead via V2X; and it obtains cloud-based fog warnings and regional perception resource distribution via 5G / Uu. A sliding time window algorithm is used to update fog concentration, road conditions, and other data in real time.

[0047] Dynamic scene cognition: The environmental interference level gradually escalates from mild interference to severe interference (visibility drops from 300 meters to 150 meters), the occlusion level is moderate occlusion, the danger object level is an emergency danger target (the vehicle in front suddenly brakes, TTC < 2 seconds), and the scene risk type is an active safety hazard scene and a passive perception limited scene.

[0048] Two-way intelligent trigger decision: (1) Active sharing and distribution triggering: Due to the emergency dangerous target, the hierarchical sharing conditions are met and the CBR is low, active sharing and distribution is triggered. The time window is 200ms, and messages (including the front vehicle ID, location, speed, danger level, event type and fog concentration change information) are sent to the following vehicles, RSU and the cloud.

[0049] (2) Targeted sharing request triggered by the receiving end: Due to severe interference causing perception failure, the precise request conditions are met, and the vehicle sends a targeted sharing request to the cloud and the forward RSU (request area 3km ahead, request type road environment, event perception, high priority). Subsequently, the CBR increases (>60%), and the load dynamic adaptation unit automatically adjusts the strategy, suspending non-emergency messages, retaining only incremental updates of emergency dangerous targets, and further simplifying the messages (only retaining core risk parameters).

[0050] Intelligent message scheduling and perception closed-loop update: The sending end updates incrementally in 200ms time windows; the vehicle behind brakes in advance after receiving the message; after receiving the road condition information ahead returned by the cloud, the vehicle updates the driving assistance strategy, decelerates, changes lanes to avoid the obstacle, and feeds back the road condition information to the dynamic scene cognition module to adjust the message sending frequency.

[0051] Example 3: Dynamic Scene of Sudden Light Changes at Tunnel Entrances and Exits This embodiment corresponds to dynamic interference from strong backlighting at the tunnel entrance and exit, posing a potential danger.

[0052] Multi-source data acquisition and dynamic fusion: As the vehicle exits the tunnel, it collects data on dynamic changes in light intensity and suspected obstacles ahead; it receives RSU data from the tunnel entrance via V2X and acquires road conditions around the tunnel via 5G / Uu. Kalman filtering is used to fuse the light intensity and obstacle perception data.

[0053] Dynamic scene cognition: The environmental interference level is moderate interference (strong light and backlight, reduced perception accuracy), the occlusion level is slight occlusion, the dangerous object level is dangerous target (suspected obstacle in front, confidence level is below the threshold), and the scene risk type is active safety hazard scene and passive perception limited scene.

[0054] Two-way intelligent trigger decision: (1) Active sharing and distribution triggering: When a dangerous target meets the hierarchical sharing conditions, active sharing and distribution is triggered. The time window is 500ms. Messages (including obstacle ID, location, type, danger level, confidence level and light change information) are sent to vehicles and RSUs that are about to exit the tunnel.

[0055] (2) Receiver-side directional sharing request trigger: Due to low perception confidence, the precise request condition is met, and this vehicle sends a directional sharing request to the RSU (request area: obstacle location area; request type: precise obstacle identification and size measurement; priority: medium). CBR is low, so it is sent normally.

[0056] Intelligent message scheduling and perception closed-loop update: The sending end updates incrementally in 500ms time windows; after the vehicle receives accurate data returned by the RSU (confirming a large rock ahead), it updates the danger level to an emergency danger target, adjusts the message priority to the highest and the time window to 300ms, and sends a more urgent warning to the surrounding area; the RSU uploads the information to the cloud.

[0057] Example 4: Dynamic scene of pedestrians / animals crossing a rural road This embodiment corresponds to rural roads where pedestrians and wild animals cross, and there is no complete roadside equipment support.

[0058] Multi-source data acquisition and dynamic fusion: This vehicle travels on rural roads, collecting the location, speed, and behavioral trajectories of pedestrians and wild rabbits; it receives sensor data from neighboring vehicles behind via V2V, and obtains cloud-based early warnings of frequent pedestrian / animal crossings and regional sensor resource distribution via 5G / Uu. A target tracking algorithm is used to update the trajectories of pedestrians and wild rabbits in real time.

[0059] Dynamic scene cognition: Environmental interference level: mild interference; Occlusion level: mild occlusion; Dangerous object level: urgent dangerous target (pedestrian TTC=2.8 seconds, wild rabbit TTC=4.2 seconds, and the pedestrian is judged to have no intention to avoid it); Scene risk type: active safety hazard scene and passive perception limited scene (cannot perceive hidden targets on the roadside).

[0060] Two-way intelligent trigger decision: (1) Active sharing and distribution triggering: When an emergency dangerous target meets the hierarchical sharing conditions and has a low CBR, active sharing and distribution is triggered. The time window is 300ms, and messages (including pedestrian and rabbit ID, location, speed, crossing trajectory, danger level, behavior prediction and the vehicle's avoidance intention) are sent to surrounding vehicles and the cloud.

[0061] (2) Targeted sharing request triggering at the receiving end: Due to the limitations of passive perception, if the conditions for accurate request are met, this vehicle will send targeted sharing requests to two neighboring vehicles in the cloud based on the distribution of perception resources (the request area is within 50 meters of the roadside, the request type is pedestrian and animal recognition, and the priority is high).

[0062] Intelligent message scheduling and perception closed-loop update: The sending end updates incrementally in 300ms time windows; after receiving the message, the neighboring vehicle slows down and sounds its horn, and feeds back its own perception data (confirming that there are no hidden targets on the roadside); after integrating the data, the vehicle adjusts its avoidance strategy until the pedestrian or wild rabbit leaves the road; the avoidance result is fed back to the dynamic scene cognition module, and message sending and requests are stopped; the cloud updates the warning information for this road section.

[0063] Example 5: Dynamic Fence Scene in Urban Construction Area This embodiment corresponds to temporary construction on urban roads, with dynamic adjustments to the construction barriers and random movement of construction personnel / equipment.

[0064] Multi-source data acquisition and dynamic fusion: When the vehicle is driving on a main urban road and encounters sudden construction ahead, it collects data on the location of the construction site fence, adjusts its speed, the extent of the fence, the location and movement trajectory of construction personnel and machinery, and surrounding traffic flow data. It receives temporary RSU construction warnings via V2X and obtains the overall situation of the construction area from the cloud via 5G / Uu. Kalman filtering is used to fuse the fence boundary and the target location.

[0065] Dynamic scene perception: Environmental interference level: moderate interference (evening backlight, construction machinery noise), occlusion level: moderate occlusion (fence obscuring internal details), hazard object level: construction personnel: emergency hazard target; construction machinery: hazard target, scene risk type: active safety hazard scene and passive perception limited scene.

[0066] Two-way intelligent trigger decision: (1) Active sharing and distribution triggering at the sending end: The hierarchical sharing conditions are met, CBR is medium, and the priority is set according to the hazard level (construction personnel are the highest, construction machinery is medium), with time windows of 300ms and 500ms respectively. This vehicle sends messages to the following vehicles, RSU, and the cloud (including the boundary of the construction area, the speed of the fence adjustment, the location of construction personnel / equipment, the hazard level, the location of scattered materials, etc.), and the messages of construction machinery are slightly simplified.

[0067] (2) Targeted sharing request triggering at the receiving end: Due to the limitations of passive perception, if the conditions for accurate request are met, this vehicle, in conjunction with the distribution of cloud-based perception resources, prioritizes sending targeted sharing requests to temporary RSUs and neighboring vehicles (requesting the area inside the enclosure, request type: new detection of construction personnel / equipment, confirmation of the final boundary of the enclosure, with high priority). Subsequently, as the CBR increases, the load dynamic adaptation unit temporarily suspends non-urgent updates of construction machinery and performs secondary simplification of construction personnel messages.

[0068] Intelligent message scheduling and perception closed-loop update: The sending end updates incrementally according to the corresponding time window; vehicles behind slow down and change lanes after receiving the message; after receiving the response data returned by the RSU and neighboring vehicles (confirming the final boundary of the enclosure and that there are no new personnel inside), the vehicle updates the driving risk assessment and optimizes the driving assistance strategy to pass smoothly; the traffic status is fed back to the dynamic scene cognition module to adjust the message sending frequency; the RSU broadcasts warnings in sync, the construction management unit adjusts the construction rhythm, and the cloud updates traffic guidance information.

[0069] In summary, the above five embodiments fully verify the effectiveness, innovation, and practicality of the triggering mechanism and system of the present invention. They can effectively solve the core pain points in the prior art, adapt to the safe passage requirements of L2+ and above intelligent connected vehicles, and have extremely high promotional value.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for triggering V2X perception sharing messages based on dynamic scene awareness at the vehicle end, characterized in that, The method specifically includes the following steps: S1. Vehicle-side multi-source data acquisition and dynamic fusion: Intelligent connected vehicles collect real-time traffic objects, road environment and their own operation data through on-board composite sensors. At the same time, they receive roadside perception data from roadside units through V2X communication and obtain global situational data from the cloud through 5G / Uu interface. The multi-source data is time-stamped, coordinate system unified and target association calibrated. Combined with the dynamic changes in the driving scene, Kalman filtering or sliding time window algorithm is used to realize real-time data fusion and status update, and output dynamic fused perception data. S2. Vehicle-side dynamic scene cognition: Based on the dynamic fusion perception data, a multi-dimensional scene judgment algorithm is used to output the dynamic judgment results of occlusion level, environmental interference level, dangerous object level and scene risk type in real time using a state machine model. The judgment results are updated in real time as the driving scene changes through a time series prediction method. The scene risk types include active safety hazard scenes and passive perception limited scenes. S3, Two-way intelligent trigger decision-making: S31. Active sharing and distribution triggering: When the scene recognition result determines that it is an active safety hazard scene and meets the preset hierarchical sharing conditions, the sending end is triggered to send a structured and simplified active sharing and distribution message to surrounding vehicles, roadside systems, and cloud management departments through one or more communication methods such as V2V, V2I, and V2N. The message sending priority is dynamically adjusted according to the hazard level. S32. Receiver-side directional sharing request trigger: When the scene perception result determines that it is a passive perception limited scene and meets the preset accurate request conditions, the receiver is triggered to send a directional sharing request message to vehicles, roadside units, and cloud platforms in the specified area through one or more communication methods such as V2V, V2I, and V2N, specifying the request scope and demand type, and prioritizing the node that covers the request area and has the best communication link quality as the request target. S4. Intelligent message scheduling and perception closed-loop update: The sending end performs hierarchical frequency control and lifecycle management on the proactively shared and distributed messages. After parsing and verifying the collaborative perception response data, the receiving end integrates it into the vehicle's dynamic perception model, updates the driving risk assessment and driving assistance strategies in real time, and dynamically adjusts the triggering logic based on the perception supplement results to form a closed-loop collaboration.

2. The vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition according to claim 1, characterized in that, In step S1, the vehicle-mounted composite sensor includes a high-definition camera, a multi-mode millimeter-wave radar, a lidar, an IMU inertial measurement unit, and a GNSS high-precision positioning module. The traffic object data includes the type, location, relative speed, motion trajectory, and behavior prediction information of vehicles, pedestrians, and non-motorized vehicles. The road environment data includes light intensity, raindrop / fog concentration, road surface adhesion coefficient, road signs, and temporary road condition anomaly information.

3. The vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition according to claim 2, characterized in that, In step S2, the specific process of dynamic scene cognition includes: S21. Occlusion Level Determination: Based on the vehicle's field of view, obstacle outline features, lane topology, and dynamic position information of neighboring vehicles, combined with the sensing range of onboard sensors, the type and level of blind spot occlusion, intersection occlusion, curve occlusion, and large vehicle occlusion are determined and divided into four levels: no occlusion, mild occlusion, moderate occlusion, and severe occlusion. The occlusion status changes are tracked in real time through a target tracking algorithm. S22. Environmental Interference Level Determination: Based on light intensity, raindrop / fog concentration, road surface adhesion coefficient, camera imaging quality, and electromagnetic interference intensity, it is divided into four levels: no interference, mild interference, moderate interference, and severe interference, with a focus on distinguishing between dynamic interference and static interference. S23. Hazardous Object Level Determination: Based on target type, relative speed, distance, collision time (TTC), lane conflict probability, and behavior prediction results, a dynamic risk assessment algorithm classifies targets into four levels: general targets, targets of concern, hazardous targets, and emergency hazardous targets. Emergency hazardous targets are further assessed using real-time trajectory prediction parameters. The behavior prediction is based on an intent recognition model that considers the relationship between the target's historical trajectory and lane lines, or on a conflict detection algorithm that considers the target's motion state and road topology. S24. Scenario Risk Type Determination: When the level of dangerous object is greater than or equal to the target of concern, or when a sudden traffic incident is identified as posing a potential threat to surrounding vehicles, it is determined to be an active safety hazard scenario; when the level of occlusion is greater than or equal to moderate occlusion, or the level of environmental interference is greater than or equal to moderate interference leading to a decrease in perception accuracy, or when driving on road sections such as curves, tunnels, and intersections that are prone to generating blind spots in perception and cannot obtain complete perception information, it is determined to be a passive perception-restricted scenario. The two types of scenarios can coexist and can be switched in real time through a state machine.

4. The vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition according to claim 3, characterized in that, In step S31, the hierarchical sharing conditions are: the level of dangerous object ≥ dangerous target; or, the occlusion level ≥ moderate occlusion and the environmental interference level ≥ moderate interference; or, an accident, construction, road obstacles, road collapse, pedestrian / animal crossing, or other sudden events are identified as posing a potential threat to vehicles approaching from behind; the active sharing and distribution message is structured and simplified data, which at least includes target ID, real-time location, speed, heading, type, danger level, confidence level, event type, occurrence location, and risk duration information, with different event types supplemented with exclusive feature parameters.

5. The vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition according to claim 4, characterized in that, In step S32, the precise request conditions are: occlusion level ≥ moderate occlusion and the sensor itself cannot fill the blind spot; or, the perception confidence of a specific target is lower than a preset threshold and cannot be calibrated by its own data; or, the environmental interference level ≥ severe interference causing perception failure; or, driving to road sections such as curves, tunnels, and intersections that are prone to perception blind spots and needing to obtain beyond-line-of-sight information; the directional sharing request message carries at least the coordinates of the request area, request type, priority, effective time limit, and a description of its own perception shortcomings. The request type includes traffic object recognition, event detection, road environment perception, and beyond-line-of-sight road condition query.

6. The vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition according to claim 5, characterized in that, In step S4, the hierarchical frequency control and lifecycle management specifically involves dividing time windows according to the level of the hazardous object: the time window for emergency hazardous targets is 100ms-300ms, the time window for hazardous targets is 300ms-500ms, and the time window for targets of concern is 500ms-1000ms. The same target or event is sent only once within the corresponding time window, and expired messages are automatically discarded to avoid message redundancy caused by repeated sending.

7. The vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition according to claim 6, characterized in that, In step S5, the sending of both the proactive sharing and distribution message and the targeted sharing request message are dynamically adapted to the current V2X communication load, which is measured by the Channel Busy Rate (CBR). When the CBR is lower than a preset threshold, all types of messages that meet the conditions are sent normally. When the CBR is higher than the preset threshold, messages related to urgent and dangerous targets are sent first, while non-urgent messages are temporarily suspended. At the same time, the messages are simplified a second time. The second simplification includes removing historical trajectory points, removing low-confidence environmental interference information, or aggregating multiple related targets into a group of information for transmission.

8. A vehicle-side V2X perception sharing message triggering system based on dynamic scene cognition, used to implement the method of any one of claims 1-7, characterized in that, The system includes: Multi-source dynamic fusion module: Composed of vehicle-mounted composite sensor interface, roadside data receiving interface, cloud data receiving interface and dynamic fusion unit, it is used to collect vehicle perception data, receive roadside and cloud data, complete spatiotemporal alignment, target association calibration and dynamic fusion of multi-source data, output dynamic fused perception data, and support sensor fault adaptive switching. Dynamic Scene Recognition Module: Connected to the multi-source dynamic fusion module, it has built-in occlusion level determination unit, environmental interference level determination unit, dangerous object level determination unit and scene risk type determination unit. It adopts a multi-dimensional dynamic judgment algorithm and state machine model to output scene recognition results in real time and update them synchronously. It also has a built-in scene prediction unit. Based on historical perception data and real-time dynamic fusion data, it uses time series prediction method to predict the trend of scene risk changes and trigger message sending or request preparation in advance to shorten the triggering delay. The bidirectional intelligent triggering module, connected to the dynamic scene cognition module, includes a sending-end hierarchical triggering unit, a receiving-end directional triggering unit, and a load dynamic adaptation unit. The sending-end hierarchical triggering unit triggers message sending according to the danger level based on the scene cognition results and hierarchical sharing conditions. The receiving-end directional triggering unit initiates directional requests based on the scene cognition results and precise request conditions, and optimizes the request object based on the regional perception resource distribution information distributed from the cloud. The load dynamic adaptation unit monitors the communication load in real time and dynamically adjusts the message sending priority, sending timing, and message simplification level according to the load status. V2X intelligent communication module: connected to the bidirectional intelligent triggering module, supporting multiple communication modes such as V2V, V2I, and V2N, used to send active sharing and distribution messages and targeted sharing request messages, receive collaborative sensing response data, complete message protocol parsing, link monitoring, anomaly handling, and secondary message simplification to ensure transmission stability and real-time performance; Message scheduling and perception closed-loop module: connected to the V2X intelligent communication module, including a hierarchical frequency control unit, a lifecycle management unit, and a dynamic perception update unit. The hierarchical frequency control unit and the lifecycle management unit realize hierarchical message scheduling and expiration handling. The dynamic perception update unit integrates the received response data into the vehicle's dynamic perception model, updates the risk assessment and driving assistance strategies, and feeds back the perception update results to the dynamic scene cognition module and the bidirectional intelligent triggering module to adjust the scene cognition accuracy and triggering logic, forming a collaborative closed loop.

9. The vehicle-side V2X perception sharing message triggering system based on dynamic scene cognition according to claim 8, characterized in that, The load dynamic adaptation unit presets a low threshold and a high threshold for CBR. When CBR is lower than the low threshold, it is determined to be low load and normal transmission is performed. When CBR is higher than the high threshold, it is determined to be high load, triggering secondary simplification and suspending non-urgent messages. When CBR is between the two, it is determined to be medium load and only low-priority messages are lightly simplified.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the vehicle-side V2X perception sharing message triggering method based on dynamic scene cognition as described in any one of claims 1-7.

Images