Commercial vehicle data closed-loop system and control method thereof
By integrating AEBS, driving recorder and DMS into a single hardware platform in commercial vehicles and building a data closed loop from vehicle to cloud, the problems of high system cost and difficulty in iteration are solved, and the continuous evolution and adaptation to complex working conditions of the system are realized.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-25
- Publication Date
- 2026-04-14
AI Technical Summary
In existing commercial vehicle systems, the AEBS system and driving recorder functions are independent, resulting in the inability to share hardware resources, high system costs, difficulty in dealing with complex road conditions, lack of data closed-loop link, and inability to achieve system iteration and optimization.
By integrating AEBS, driving recorder and DMS into a single hardware platform, a data closed loop from vehicle to cloud is built. Multi-source information is fused through image processor SOC and microcontroller MCU to generate control commands, which are then encrypted and stored in a protected memory. High-value data is selected and uploaded to the cloud for model iteration.
It achieves hardware resource sharing, reduces system costs, builds a complete data closed loop, enables the system to continuously evolve, adapt to complex working conditions, and provides multi-dimensional data support for new business models.
Smart Images
Figure CN121849166A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent connected vehicle technology, and in particular to an integrated data closed-loop system for commercial vehicles that integrates regulatory requirements, active safety and driver monitoring, and its control method. Background Technology
[0002] With increasingly stringent regulations and the development of intelligent driving technology, commercial vehicles, especially heavy trucks, are facing multiple mandatory requirements. For example, China's Ministry of Transport has passed relevant standards requiring commercial freight vehicles to be equipped with Automatic Emergency Braking Systems (AEBS) as standard, and requiring vehicles weighing over 12 tons to be equipped with driving recorders with video recording capabilities.
[0003] To meet the above requirements, the industry currently generally adopts two independent system solutions: an "AEBS forward-looking integrated system" and a "driving recorder with national standards." This solution has significant drawbacks: First, hardware, storage, and communication resources cannot be shared, resulting in high system costs and hindering cost control for commercial vehicles, which are used as means of production. Second, the decision-making logic of the AEBS system is based on preset rules for limited scenarios, making it difficult to cope with complex real-world road conditions. Furthermore, the massive amounts of perception and decision-making data generated by the system are stored on the vehicle side due to the lack of an effective data closed-loop link, making it unusable for system iteration and optimization.
[0004] In addition, some existing technologies focus on remote driving control systems, which achieve remote vehicle control by deploying a large number of high-cost sensors and placing the main computing power in the cloud. These solutions are usually expensive and do not integrate with functions such as driving recorders required by regulations. Other solutions focus on human-machine interaction and driver status monitoring in the cockpit, but do not involve deep integration and data closure with vehicle active safety control systems (such as AEBS).
[0005] Therefore, there is an urgent need for an integrated commercial vehicle solution that can meet regulatory requirements, optimize costs, and build an effective data loop to drive the continuous evolution of intelligent driving capabilities. Summary of the Invention
[0006] The purpose of this invention is to provide a data closed-loop system for commercial vehicles and its control method, in order to solve the problem of how to achieve integrated system cost optimization and continuous and efficient iteration of intelligent driving capabilities through hardware integration and data closed-loop collaborative design, while meeting mandatory regulatory requirements.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a commercial vehicle data closed-loop system, comprising: Integrated ECU; The ADAS camera, DMS camera, millimeter-wave radar, protection memory, and drive-by-wire chassis actuator are connected to the integrated ECU. The integrated ECU includes an image processor SOC, a microcontroller MCU, and a wireless communication module; The image processor SOC is used to process images from ADAS cameras and DMS cameras and output perception information. The microcontroller MCU is used to fuse the perception information of the image processor SOC, the millimeter-wave radar information and the vehicle CAN information to generate AEBS control commands and send them to the drive-by-wire chassis actuator. The image processor SOC is also used to encrypt and store ADAS video data, DMS video data, radar data, AEBS control commands and vehicle CAN data in the protective memory, and upload the data to the cloud platform through the wireless communication module after filtering the data based on preset scene priority rules; in the scene priority rules, the priority of each scene from high to low is as follows: accident scene, AEBS braking scene, forward collision warning (FCW) scene, and normal scene.
[0008] In one embodiment, the integrated ECU further includes a positioning module and a backup battery module; the backup battery module is used to provide power for data storage and uploading when the vehicle loses power.
[0009] In one embodiment, the protective memory supports high-temperature fire protection, ember protection, water immersion protection, and pressure resistance protection.
[0010] In one embodiment, the wireless communication module is in a low-power mode when the vehicle's ACC is off, and can be woken up by a remote wake-up command from the cloud platform, thereby waking up the integrated ECU.
[0011] In one embodiment, the image processor SOC and the microcontroller MCU communicate with dual redundancy via SPI and UART interfaces.
[0012] Secondly, this invention provides a closed-loop control method for commercial vehicle data, implemented based on the above-mentioned system, comprising the following steps: Data acquisition, processing and local storage steps: Raw data is acquired through ADAS camera, DMS camera and millimeter wave radar; the image processor SOC identifies and processes the ADAS video stream and DMS video stream from the camera, and the microcontroller MCU fuses multi-source information to perform AEBS decision control. At the same time, the raw data and the control instructions and status data in the AEBS decision process are encrypted and stored in a protective memory. High-value data screening and preprocessing steps: The image processor SOC performs local screening of the data stored in the protection memory based on preset scene priority rules. In the scene priority rules, the priority of each scene from high to low is as follows: accident scene, AEBS braking scene, forward collision warning (FCW) scene, and normal scene; the screened data is preprocessed by timestamp synchronization and coordinate system alignment. Data encapsulation and uploading steps: The preprocessed data is associated with the encapsulation environment and vehicle status tags, and the data is desensitized and compressed. The ADAS video stream and DMS video stream are video encoded and compressed to form a data packet, which is then uploaded to the cloud platform through the wireless communication module. Cloud service and model iteration steps: The cloud platform uses the uploaded data packet to perform data service and generates a model upgrade instruction based on the service result; in response to the model upgrade instruction, an OTA upgrade package is sent to the integrated ECU through the wireless communication module to update the algorithm model in the integrated ECU.
[0013] In one embodiment, in the high-value data screening and preprocessing step, a lightweight neural network model is used to pre-screen key frames in the ADAS video stream and DMS video stream, the key frames including target mutation frames or trajectory change frames; the lightweight neural network model is YOLOv5n or MobileNetV3.
[0014] In one embodiment, during the data encapsulation and uploading step, the associated environmental tags include weather information, light intensity, and road hazard index obtained based on map services.
[0015] In one embodiment, in the data encapsulation and uploading step, the compression and packaging involves merging the ADAS video stream, DMS video stream, radar point cloud data, and vehicle CAN bus data within a preset time window before and after the triggering event, after spatiotemporal alignment.
[0016] In one embodiment, in the cloud service and model iteration step, the data service includes: using the DMS video stream in the data packet to perform secondary identification and analysis of the driver's risk status, and / or fusing AEBS event data, DMS status data and vehicle diagnostic data in the data packet to perform insurance risk control modeling, so as to generate risk assessment results.
[0017] Beneficial effects Compared with the prior art, the present invention has at least the following beneficial effects: 1. This invention integrates the driving recorder, AEBS, and DMS into a single hardware platform, achieving the sharing of hardware resources, storage resources, and communication resources, and significantly reducing system costs and installation complexity.
[0018] 2. This invention constructs a complete data closed loop from the vehicle to the cloud, enabling the collection, filtering and use of massive amounts of real-world scenario data for algorithm training. This completely changes the limitations of traditional AEBS based on fixed rules, allowing the system to continuously evolve and adapt to complex working conditions.
[0019] 3. The multi-dimensional data (ADAS+DMS+vehicle body+positioning) integrated in this invention provides a reliable data foundation for new business models such as "technology insurance claims reduction", "driver behavior rating", and "dynamic management of operational risks".
[0020] 4. This invention establishes a multi-layered data security and system reliability assurance system by employing independent protective storage, backup batteries, dual redundant communication, and an accident marking mechanism. This ensures the integrity of critical data under any accident conditions and supports remote silent upgrades, thereby improving system maintainability. Attached Figure Description
[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 A schematic diagram of the hardware architecture of a commercial vehicle data closed-loop system provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the closed-loop data control method for commercial vehicles provided in an embodiment of the present invention. Detailed Implementation
[0023] The embodiments of this application will now be described in detail with reference to the accompanying drawings.
[0024] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0025] Example 1: like Figure 1 As shown, this embodiment of the invention provides a closed-loop data system for commercial vehicles. The core of the system is an integrated ECU; The integrated ECU is a hardware controller that includes: a SOC image processing module, an MCU module, a SerDes video serial module, an AHD / SerDes video serial module, a Bluetooth / WIFI module, a built-in SD / Flash storage module, a TFT LCD screen, a wireless communication module (such as a 4G / 5G communication module), a USB / SDIO communication module, a CAN communication module, a driver IC card, a button panel, a positioning module, a backup battery module, an acceleration sensor module, a pluggable SD card, and a read / write USB interface.
[0026] The external key components connected to the integrated ECU (100) include: Sensors: ADAS cameras, DMS cameras, millimeter-wave radar; Data security and execution unit: independent protected storage, drive-by-wire chassis actuators (such as EBS system) Interactive unit: Central control multimedia display unit.
[0027] The system communicates with a remote cloud platform via a wireless communication module within the integrated ECU and its external antenna (such as a 4G / 5G network communication antenna). The cloud platform includes a back-end freight platform and an enterprise platform.
[0028] The SOC image processing module is connected to the built-in SD / Flash storage module. It connects to the MCU module via SPI, UART, and GPIO interfaces, forming a dual-redundant communication channel through one SPI interface and one UART interface, ensuring the reliability of critical sensing and control command transmission. The SOC image processing module also connects to the TFT LCD panel via SPI, to the Bluetooth / WIFI module via UART and SDIO interfaces, to the SerDes video serial module via MIPI, to the AHD / SerDes video serial module via BT1120 / CSI, and to the wireless communication module (i.e., the 4G / 5G communication module) via IIS; it also connects to the USB / SDIO communication module.
[0029] The MCU module connects to the CAN communication module, the driver's IC card via I2C, the button panel via GPIO, the positioning module via UART, the battery backup module via ADC, and the acceleration sensor module via I2C. The MCU module also connects to an external main power supply and a vehicle signal detection unit (used to detect signals such as left / right turn, reverse gear, and vehicle speed) via ADC and GPIO interfaces.
[0030] The USB / SDIO communication module has a pluggable SD card and a read / write USB interface.
[0031] At the sensor and input level: The ADAS camera is connected to the Serdes video serial module in the integrated ECU via an LVDS high-definition interface, and is used to collect high-definition video streams of the road in front of the vehicle in real time.
[0032] The DMS camera is connected to the AHD / Serdes video serial module in the integrated ECU via an LVDS or AHD high-definition interface. It is used to collect video streams of the driver's face and the cabin, and integrates audio signals collected by the MIC.
[0033] The millimeter-wave radar is connected to the CAN communication module in the integrated ECU via the CAN1 bus to provide information on the distance, speed, and angle of the target ahead.
[0034] The protective memory, as an independent external secure storage unit, is connected to the SOC image processing module in the integrated ECU via the SDIO interface, and is used to securely store all critical data.
[0035] In this embodiment of the invention, the positioning module (such as a GPS / BeiDou module) and its external positioning antenna are used to provide real-time vehicle location, speed and time information.
[0036] The backup battery module powers data storage and uploading when the vehicle loses power. Specifically, when the vehicle's main power supply is unexpectedly disconnected, the backup battery module provides continuous power to the system, ensuring secure data storage and the completion of emergency communications.
[0037] The wireless communication module (4G / 5G communication module) is responsible for data interaction with the cloud platform.
[0038] At the execution and output level: The chassis drive-by-wire actuator (such as the EBS electronic braking system) is connected to the CAN communication module of the integrated ECU via the CAN2 bus to receive and execute braking commands.
[0039] The central control multimedia display unit includes the original vehicle central control multimedia display screen and the central control display screen. The original vehicle central control multimedia display screen is connected to the CAN communication module in the integrated ECU via the CAN3 bus for command interaction. The central control display screen is connected to the AHD / Serdes video serial module in the integrated ECU via the AHD / LVDS video interface for playback of driving record video and local retrieval of accident video data stored in SD.
[0040] In this embodiment of the invention, the SOC image processing module includes an image processor SOC, and the MCU module includes a microcontroller MCU. The image processor SOC processes images from the ADAS camera and the DMS camera and outputs perception information; the microcontroller MCU fuses the perception information from the image processor SOC, millimeter-wave radar information, and vehicle CAN information to generate AEBS control commands and send them to the chassis drive-by-wire actuator. The image processor SOC and the microcontroller MCU communicate with each other via SPI and UART interfaces with dual redundancy to ensure the reliability of control command transmission.
[0041] The basic workflow of the system is as follows: The image processor SOC performs real-time decoding and deep learning inference on ADAS and DMS video streams from ADAS and DMS cameras. For the ADAS video stream, it identifies lane lines, vehicles, pedestrians, traffic signs, cones, and other targets and their location information. For the DMS video stream, it identifies the driver's facial features, eyelid state, gaze direction, and held objects (such as mobile phones or cigarettes), making a preliminary judgment on fatigue and distraction. This structured perception information is transmitted to the MCU via the SPI / UART bus.
[0042] The microcontroller (MCU) simultaneously receives a target list from millimeter-wave radar and vehicle CAN information such as vehicle speed, accelerator pedal travel, brake pedal status, and steering wheel angle from the vehicle's CAN network. The MCU fuses this multi-source information, performs tracking and prediction, and ultimately runs the AEBS decision-making algorithm. When the system determines there is a collision risk, the MCU generates AEBS control commands and status data, including warning level and required deceleration, and sends them to the chassis drive-by-wire actuators via the CAN2 bus to achieve automatic emergency braking of the vehicle.
[0043] Meanwhile, the image processor SOC, acting as a central data recording unit, performs data acquisition, processing, and local storage. It synchronously writes all the following data in encrypted form to a protected memory and a removable SD card: raw video data from ADAS and DMS; raw or pre-processed data from millimeter-wave radar; AEBS control commands and decision-making status data issued by the MCU; and vehicle CAN data collected from the CAN network (such as vehicle speed, throttle, brakes, turn signals, wiper status, etc.). The data is then filtered based on preset scene priority rules and uploaded to the cloud platform via the wireless communication module. In the preset scene priority rules, the priorities of each scene, from highest to lowest, are: accident scene, AEBS braking scene, forward collision warning (FCW) scene, and normal scene.
[0044] In this embodiment of the invention, the protective memory is specially packaged to support high-temperature fire protection, ember protection, water immersion protection and pressure protection, ensuring that the data is still readable after a serious accident.
[0045] In this embodiment of the invention, the acceleration sensor module integrated within the integrated ECU is an accelerometer-gyroscope (IMU). This module is used to monitor the vehicle's longitudinal and lateral acceleration and yaw rate in real time. Its data is used for: first, redundancy verification with the positioning module and the vehicle's CAN bus speed signal to improve the reliability of vehicle trajectory recording; second, when a severe impact acceleration exceeding a threshold (e.g., for collision marking) or continuous abnormal lateral acceleration (e.g., for rollover risk marking) is detected, the current data segment is automatically marked as the highest priority accident scenario, and an emergency data saving process is triggered.
[0046] Example 2 like Figure 2 As shown, this embodiment of the invention also provides a closed-loop control method for commercial vehicle data, implemented based on the above system. This method is a continuous, cyclical "vehicle-cloud" collaborative process, specifically including the following steps: S100: Data Acquisition, Processing, and Local Storage After the system is powered on, it continues to execute the workflow described in Example 1. That is, it acquires raw data through the ADAS camera, DMS camera, and millimeter-wave radar; the image processor SOC processes and identifies the ADAS video stream and DMS video stream from the cameras; and the microcontroller MCU fuses the multi-source information for AEBS decision-making and control. Simultaneously, the raw data, control commands, and status data from the AEBS decision-making process are encrypted and stored in a protected memory. This step provides the data foundation for a closed-loop data system.
[0047] S200: High-Value Data Screening and Preprocessing This step is executed periodically in the background by the image processor SOC or triggered by events. Specifically, it includes: Filtering: The image processor SOC intelligently filters massive amounts of stored data based on a set of preset scene priority rules. These rules define the priority of each scene from highest to lowest as follows: accident scenes, AEBS braking scenes, forward collision warning (FCW) scenes, and normal scenes. For example, when the accelerometer detects a severe collision (accident scene), or the MCU records a complete AEB braking event (AEBS braking scene), the relevant data will be marked as the highest priority.
[0048] Pre-screening: To further improve efficiency, the system uses a lightweight neural network model deployed on the SOC to quickly scan the ADAS and DMS video streams during screening, pre-screening keyframes containing target mutation frames or trajectory change frames. Only video clips containing these keyframes will enter the high-priority candidate queue. This process also removes a large amount of unchanged, invalid, or duplicate data. For example, the lightweight neural network model is a YOLOv5n or MobileNetV3 model.
[0049] Preprocessing: For the selected high-value data segments, timestamp synchronization and coordinate system alignment are performed. For example, the timestamps of ADAS cameras, radar, and vehicle CAN bus are unified to the same time base; the target coordinates detected by radar are transformed into the pixel coordinate system of the camera images to facilitate subsequent fusion analysis.
[0050] S300: Data Encapsulation and Upload This step packages the preprocessed data into a standard data package that can be used for cloud analysis. Specifically, it includes: Encapsulation Tags: These tags associate the data packet with environmental and vehicle status labels. The system uses GPS information obtained through the positioning module. When the vehicle has a network connection, it requests environmental labels such as weather information, light intensity, and road hazard index for that location from a cloud-based map service. Simultaneously, the vehicle sensor status during the event (such as steering wheel angle, accelerator / brake opening, and light / wiper status) is encapsulated as part of the vehicle status tag.
[0051] Compliance and Compression: First, the video data undergoes data anonymization processing, and algorithms are used to blur the license plates of other vehicles and the faces of pedestrians appearing in the video in real time. Then, the ADAS video stream and DMS video stream are encoded and compressed using standards such as H.264 / H.265. Finally, compression and packaging are performed. The core of this process is to temporally and spatially align and merge the ADAS video stream, DMS video stream, radar point cloud data, and vehicle CAN bus data within a preset time window before and after the triggering event (e.g., 10 seconds before the event and 5 seconds after the event) to form a compact data packet containing multimodal data.
[0052] Upload: The packaged data packet is uploaded to the cloud platform via the wireless communication module. The wireless communication module is in low-power mode when the vehicle's ACC is off, but it can be woken up remotely by the cloud platform via the wireless communication network. This, in turn, wakes up the MCU module via the GPIO interface, thereby waking up the entire integrated ECU, enabling the reception or upgrade of data acquisition commands in silent mode.
[0053] S400: Cloud Services and Model Iteration After receiving the data packet (which it can preprocess by performing necessary data entry management and cleaning), the cloud platform initiates a data value mining and system evolution cycle. Specifically, this includes: Data Services: The cloud platform leverages its powerful computing capabilities to perform more complex data services. For example, it uses the DMS video stream in the data package for secondary identification and analysis of driver risk status, employing a larger neural network model to perform more refined classification of fatigue and distraction, identifying micro-fatigue states that may be missed by the vehicle. It integrates AEBS event data, DMS status data, and vehicle diagnostic data from the data package to perform insurance risk control modeling and generate risk assessment results. For instance, the platform analyzes the AEBS trigger frequency of a fleet or a driver, the driver's state at the time of triggering (whether fatigued), and the vehicle's rapid acceleration and deceleration behavior to generate a personalized risk score. This score can be directly used by insurance companies for "technology-driven claims reduction" or for fleet safety performance evaluation.
[0054] Model training and iteration: The cloud aggregates massive amounts of similar scene data uploaded by vehicles, and professionals annotate the data to retrain or optimize the AEBS perception model, decision-making algorithm, or DMS recognition model.
[0055] OTA Upgrade: Based on the new model or optimized parameters obtained from training, the cloud platform generates a model upgrade command and creates an OTA upgrade package. Subsequently, in response to the model upgrade command, the OTA upgrade package is sent to the integrated ECUs of eligible vehicles in the fleet via wireless communication modules. After receiving and verifying the upgrade package under safe conditions (e.g., parked, powered on), the vehicles update the algorithm models (e.g., the target detection model in the SOC, the decision rule parameters in the MCU) within the integrated ECU. After the update is completed, the vehicles have a stronger ability to cope with corresponding scenarios, thus realizing a complete closed loop from real-world data collection to system capability improvement.
[0056] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A commercial vehicle data closed-loop system, characterized in that, include: Integrated ECU; The ADAS camera, DMS camera, millimeter-wave radar, protection memory, and drive-by-wire chassis actuator are connected to the integrated ECU. The integrated ECU includes an image processor SOC, a microcontroller MCU, and a wireless communication module; The image processor SOC is used to process images from ADAS cameras and DMS cameras and output perception information. The microcontroller MCU is used to fuse the perception information of the image processor SOC, the millimeter-wave radar information and the vehicle CAN information to generate AEBS control commands and send them to the drive-by-wire chassis actuator. The image processor SOC is also used to encrypt and store ADAS video data, DMS video data, radar data, AEBS control commands and vehicle CAN data in the protective memory, and upload the data to the cloud platform through the wireless communication module after filtering the data based on preset scene priority rules; in the scene priority rules, the priority of each scene from high to low is as follows: accident scene, AEBS braking scene, forward collision warning (FCW) scene, and normal scene.
2. The commercial vehicle data closed-loop system according to claim 1, characterized in that, The integrated ECU also includes a positioning module and a backup battery module; the backup battery module is used to provide power for data storage and uploading when the vehicle loses power.
3. The commercial vehicle data closed-loop system according to claim 1 or 2, characterized in that, The protective storage device supports high-temperature fire protection, ember protection, water immersion protection, and pressure resistance protection.
4. The commercial vehicle data closed-loop system according to claim 1, characterized in that, The wireless communication module is in a low-power mode when the vehicle's ACC is off, and can be woken up by a remote wake-up command from the cloud platform, thereby waking up the integrated ECU.
5. The commercial vehicle data closed-loop system according to claim 1, characterized in that, The image processor SOC and the microcontroller MCU communicate with each other via dual redundancy through SPI and UART interfaces.
6. A closed-loop control method for commercial vehicle data, implemented based on the system described in any one of claims 1-5, characterized in that, include: Data acquisition, processing and local storage steps: Raw data is acquired through ADAS cameras, DMS cameras and millimeter-wave radar; The image processor SOC identifies and processes the ADAS video stream and DMS video stream from the camera, and the microcontroller MCU fuses the multi-source information to perform AEBS decision control. At the same time, the original data and the control instructions and status data in the AEBS decision process are encrypted and stored in a protective memory. High-value data screening and preprocessing steps: The image processor SOC performs local screening of the data stored in the protection memory based on preset scene priority rules. In the scene priority rules, the priority of each scene from high to low is as follows: accident scene, AEBS braking scene, forward collision warning (FCW) scene, and normal scene; the screened data is preprocessed by timestamp synchronization and coordinate system alignment. Data encapsulation and uploading steps: The preprocessed data is associated with the encapsulation environment and vehicle status tags, and the data is desensitized and compressed. The ADAS video stream and DMS video stream are video encoded and compressed to form a data packet, which is then uploaded to the cloud platform through the wireless communication module. Cloud service and model iteration steps: The cloud platform uses the uploaded data packet to perform data service and generates a model upgrade instruction based on the service result; in response to the model upgrade instruction, an OTA upgrade package is sent to the integrated ECU through the wireless communication module to update the algorithm model in the integrated ECU.
7. The method according to claim 6, characterized in that, In the high-value data screening and preprocessing step, a lightweight neural network model is used to pre-screen key frames in the ADAS video stream and DMS video stream. The key frames include target mutation frames or trajectory change frames. The lightweight neural network model is YOLOv5n or MobileNetV3.
8. The method according to claim 6, characterized in that, In the data encapsulation and uploading steps, the associated environmental tags include weather information, light intensity, and road hazard index obtained based on map services.
9. The method according to claim 6, characterized in that, In the data encapsulation and uploading steps, the compression and packaging involves merging the ADAS video stream, DMS video stream, radar point cloud data, and vehicle CAN bus data within a preset time window before and after the triggering event.
10. The method according to claim 6, characterized in that, In the cloud service and model iteration steps, the data service includes: using the DMS video stream in the data packet to perform secondary identification and analysis of the driver's risk status, and / or fusing AEBS event data, DMS status data and vehicle diagnostic data in the data packet to perform insurance risk control modeling, so as to generate risk assessment results.