Intelligent networked automobile sensing module diagnosis system and method thereof

Abstract

The invention mainly relates to the technical field of intelligent networked automobile sensing module diagnosis, and provides a sensing module self-diagnosis system scheme, which comprises a data acquisition layer, a cloud analysis layer, a decision execution layer and a maintenance feedback layer which are in data interaction in sequence. Aiming at specific engineering problems such as early prediction of performance degradation of intelligent network connection automobile sensors and multi-sensor cooperative failure diagnosis, the method is based on a Lambert-Beer law laser radar attenuation model, traditional passive diagnosis is broken through, a predictive diagnosis method is realized, a cooperative diagnosis mechanism can be realized, multi-sensor fusion mutual verification is realized, and the method is suitable for large-scale popularization and application. And finally, through dynamic fault feedback, iterative updating and continuous optimization are realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicle perception module diagnostic technology, and more specifically to an intelligent connected vehicle perception module diagnostic system and method. Background Technology

[0002] Current environmental perception systems for intelligent connected vehicles generally adopt a multimodal sensor fusion architecture, mainly including core components such as cameras, millimeter-wave radar, lidar, ultrasonic sensors, and positioning systems (GNSS / IMU). These sensors are susceptible to weather interference, hardware aging, or software malfunctions in complex operating environments, leading to perception distortion or even failure.

[0003] However, existing industry data shows that over 70% of Level 2 and above autonomous driving accidents originate from failures in the vehicle's perception system. Traditional diagnostic methods, however, have significant limitations. They cannot detect hidden faults such as software malfunctions, protocol failures, ping mismatches, and network address discrepancies. They primarily rely on passive fault code detection (OBD-II standard), lacking the ability to predict sensor performance degradation and distortion. Furthermore, their diagnostic granularity is coarse, making it difficult to pinpoint hidden faults related to multi-sensor coordination discrepancies. Additionally, the diagnostic cycle is long, requiring vehicles to be returned to the factory for connection to specialized equipment.

[0004] Therefore, there is an urgent need to design an intelligent online diagnostic system that can monitor the collaborative status of multiple sensors in real time, accurately identify hidden problems such as software faults, protocol anomalies, and network matching deviations, and at the same time have the ability to predict the trend of perception performance degradation. This system can achieve refined and proactive fault warning and location without the need for factory return, thereby reducing the risk of autonomous driving accidents caused by perception system failures from the source. Summary of the Invention

[0005] This invention provides a diagnostic system and method for the perception module of intelligent connected vehicles to solve the problems existing in the prior art.

[0006] To achieve the above objectives, embodiments of the present invention provide a diagnostic system for a perception module of an intelligent connected vehicle, comprising: a data acquisition layer (vehicle edge layer), a cloud analysis layer, a decision execution layer, and a maintenance feedback layer that interact sequentially;

[0007] The data acquisition layer, deployed in the vehicle-mounted edge computing unit, is used to acquire raw data from cameras, lidar, millimeter-wave radar, ultrasonic sensors, and GNSS / IMU in real time, and to perform time synchronization, coordinate system transformation, feature extraction, and data compression.

[0008] The cloud-based analysis layer receives feature vectors uploaded from the edge layer and runs a multimodal diagnostic model group, including an anomaly detection model, a health prediction model, and a collaborative verification module.

[0009] The decision-making and execution layer generates maintenance guidance strategies based on fault levels, supporting AR visualization and voice interaction;

[0010] The maintenance feedback layer collects maintenance result data and dynamically updates the diagnostic rule base.

[0011] Preferably, the data acquisition layer includes vehicle startup, a multi-sensor data acquisition unit, and data preprocessing.

[0012] Preferably, the multi-sensor data acquisition device includes a camera, lidar, millimeter-wave radar, ultrasonic radar, and GUSS / IMU; the multi-sensor data acquisition device includes time synchronization, coordinate transformation, feature extraction, and data compression.

[0013] Preferably, the cloud analysis layer includes hardware fault detection, degradation mode, safe parking, feature parameter extraction, cloud data transmission, multi-level cloud data analysis, dynamic decision-making, and fault level judgment.

[0014] Preferably, the hardware fault detection includes a rule engine, hardware fault code detection, and response time.

[0015] Preferably, the cloud-based multi-level data analysis includes lifespan prediction, health assessment, and cross-sensor verification.

[0016] Preferably, the fault level determination includes recording performance degradation, scheduling maintenance reminders, and vehicle speed limit control.

[0017] Preferably, the decision execution layer includes maintenance guidance generation, AR maintenance guidance, and voice interaction.

[0018] Preferably, the maintenance feedback layer includes maintenance result feedback, collection of maintenance feedback data, and updating of the diagnostic rule base.

[0019] The present invention further discloses a diagnostic method for the perception module of an intelligent connected vehicle, comprising the following steps:

[0020] Step S1: First, start the vehicle. After the vehicle starts, data is collected through the multi-sensor data acquisition unit.

[0021] Step S2: After the vehicle environmental perception data is collected, data preprocessing is performed. First, time synchronization is performed to solve problems such as time drift, zero drift, and benchmark calibration. Then, coordinate system transformation is performed. After completion, basic feature extraction is performed. Since the amount of high-precision map data collected is particularly large, Gini data compression is also required to enable data to flow quickly.

[0022] Step S3: After data preprocessing is completed, hardware fault detection is performed. This mainly involves hardware checks of the sensing system, including hardware and software ping value matching, CAN communication, wiring harness connection faults, hardware identification, and other fault detection. Hardware fault code detection is also performed to intelligently locate the fault point more accurately, shortening the troubleshooting time. After the troubleshooting time is shortened, the sensing system sensors are activated. The response time of the sensing system sensors is specified in 4-3 as less than 100ms to be considered good. When the iteration time is greater than 100ms, the system automatically judges that the sensor data is invalid or the data is delayed, and it is necessary to reload the relevant data of other sensors to fit all the special data required.

[0023] Step S4: When a hardware fault is detected, the system immediately enters degrade mode, safely stops the vehicle, and ensures the safety of the driver and passengers. If no fault risk is detected, the system extracts feature parameters, performs full-stack early warning and data closure, and implements warning classification based on data judgment. Different vehicle functions are implemented according to different levels. Since the functions performed by each sensor are different, multi-sensor fusion technology is used here. Five different types of sensors are used as data acquisition inputs, and the main items and feature parameters detected by each sensor are the main basis. When a sensor fails or the data is inaccurate, it is necessary to use the feature parameters of other sensors to fit parameters to compensate for the missing feature values ​​of the failed sensor and ensure safe driving.

[0024] Step S5: After the feature parameters are acquired, the data is transmitted to the cloud for multi-level data analysis, including lifespan prediction, health assessment, and cross-sensor verification.

[0025] Step S6: Based on the results of multi-level data analysis in the cloud, the vehicle attitude, vehicle speed, vehicle heading angle, vehicle chassis parameters, power system parameters and vehicle parameters are analyzed to ultimately achieve dynamic decision-making;

[0026] Step S7: After dynamic decision-making is completed, the fault level is determined, and different levels correspond to different responses. For general faults with functional degradation, performance degradation needs to be recorded and fed back to the cloud big data platform to form a closed feedback data structure. When the diagnostic prompts that repair or maintenance is needed, a repair appointment reminder is initiated to remind the customer to proceed with maintenance on time, ensuring customer safety. When the prompt indicates limited functionality or a more serious fault, the vehicle speed limit control module is activated, and its operation is downgraded; if necessary, the vehicle must be stopped.

[0027] Step S8: After determining the fault level, two functions are used: maintenance guidance generation and AR maintenance guidance, and voice interaction. Based on real-time feedback data and vehicle parameters, a clear maintenance guidance document is provided to determine the specific parts and software to be repaired or maintained. Then, AR maintenance guidance and voice interaction are introduced to help customers understand the fault point repair and how to provide feedback to professional maintenance personnel.

[0028] Step S9: After completion, feedback on the repair results is provided. The feedback mechanism will statistically analyze the information and send it to the cloud big data platform to establish a comprehensive database. In the module for collecting repair feedback data, the data will be effectively filtered and updated in a timely manner to the diagnostic rule base. The rule base will then feed the data back to the cloud for multi-level data analysis.

[0029] Compared with the prior art, the present invention has the following advantages:

[0030] This application addresses specific engineering problems such as early prediction of sensor performance degradation and multi-sensor collaborative failure diagnosis in intelligent connected vehicles. It utilizes a lidar attenuation model based on Lambert-Beer's law, as shown in the formula:

[0031]

[0032] Where A n The sensor collects the absorbance of sample n, ε n Let C be the absorbance coefficient of sample n. n Let be the concentration of sample n, and L be the distance the incident light travels in sample n. The linear relationship between the absorbance of a sample and its concentration can be measured using Lambert-Beer's law, thereby predicting the performance degradation and failure of sensing sensors in intelligent connected vehicles. This breaks away from traditional passive diagnosis, realizing a predictive diagnostic method. It also enables a collaborative diagnostic mechanism, allowing for multi-sensor fusion and mutual verification, and ultimately achieving iterative updates and continuous optimization through dynamic fault feedback.

[0033] This application focuses on three areas of innovation:

[0034] ① Predictive diagnosis: Breaking through the traditional passive fault code detection mode, it can quickly and effectively identify hidden software faults, build a sensor performance degradation model, and predict or judge sensor performance and lifespan degradation based on the decreasing trend of lidar point cloud density.

[0035] ② Collaborative diagnostic mechanism: Establish cross-sensor association rules and multi-sensor fusion technology, and perform mutual verification of target tracking consistency between forward-looking camera, LiDAR and millimeter-wave radar;

[0036] ③ Dynamic optimization capability: Combined with the diagnostic model architecture, the diagnostic model can be dynamically optimized, and a database can be established for real-time feedback.

[0037] This application constructs a new generation of diagnostic framework for sensing systems by integrating dynamic model optimization, multi-sensor collaborative diagnosis, and real-time feedback mechanisms. Its core advantage lies in transforming the traditional "post-fault handling" model into a closed-loop system of "prediction-prevention-precision maintenance." Practical verification shows that this solution can predict sensor faults 48 hours in advance, improve maintenance efficiency by more than 35%, and reduce the false alarm rate to one-third of traditional methods. Attached Figure Description

[0038] Figure 1 The attached figure is a flowchart of the present invention.

[0039] Figure 2 The attached figure is a detailed flowchart of the present invention. Detailed Implementation

[0040] Please see the appendix Figure 1-2 To achieve high-precision perception system diagnosis, it is necessary to construct a layered data processing architecture for the perception modules of intelligent connected vehicles, including edge layer, cloud layer, and interaction layer.

[0041] In the edge layer, a lightweight diagnostic agent is deployed in the onboard computing unit (ECU) to collect raw sensor data streams in real time, such as camera RAW images, LiDAR point cloud sequences, millimeter-wave radar ranging, ultrasonic ranging, and GNSS / IMU positioning and navigation information. Then, data preprocessing is performed, including time synchronization (addressing time drift, zero-point drift, and baseline calibration), coordinate system transformation (inconsistencies in positioning coordinates between LiDAR, millimeter-wave radar, cameras, and GNSS / IMU can lead to distortion of the constructed object and road, resulting in decision-making errors), and basic feature extraction (such as the SSIM structural similarity of images and the utilization rate of effective parameters like radar signal-to-noise ratio, which directly affect the performance of the perception system).

[0042] In the cloud layer, it primarily receives feature vectors and feature points uploaded from the data acquisition edge layer, and runs a multimodal diagnostic model group, including an anomaly detection model, a health prediction model, and a collaborative verification module. The anomaly detection model, based on a temporal convolutional network (TCN), identifies sensor data drift and can combine data processing to analyze data drift issues. Health prediction, based on point cloud density and clarity, simultaneously analyzes and predicts the remaining lifespan of LiDAR sensors, cameras, millimeter-wave radar, ultrasonic radar, GNSS / IMU inertial navigation, etc., providing a comprehensive judgment through multi-sensor fusion. The collaborative verification module detects latent faults through a cross-sensor association rule base (such as a camera-radar ranging consistency threshold), addressing the pain points of software mismatch and hidden hardware sub-state faults.

[0043] Table 1: Diagnostic parameters of key sensors in the sensing system

[0044]

[0045] To further optimize the above technical solution, a self-diagnostic system method for the sensing module is proposed.

[0046] This invention patent includes a data acquisition layer (vehicle edge layer), a cloud analysis layer, a decision execution layer, and a maintenance feedback layer.

[0047] To further optimize the above technical solution, the data acquisition layer includes functional modules such as vehicle start-up 1, multi-sensor data acquisition device 2, camera 2-1, lidar 2-2, millimeter-wave radar 2-3, ultrasonic radar 2-4, GUSS / IMU 2-5, data preprocessing 3, time synchronization 3-1, coordinate transformation 3-2, feature extraction 3-3, and data compression 3-4; the cloud analysis layer includes hardware fault detection 4, rule engine 4-1, hardware fault code detection 4-2, response time 4-3 <100ms, degradation mode, and safe parking 5-1. The system includes the following functional modules: feature parameter extraction (5-2), cloud data transmission (6), multi-level cloud data analysis (7), lifespan prediction (7-1), health assessment (7-2), cross-sensor verification (7-3), dynamic decision-making (8), fault level judgment (9), performance degradation recording (9-1), scheduled maintenance reminders (9-2), and vehicle speed limit control (9-3). The decision execution layer includes two functional modules: maintenance guidance generation (10) and AR maintenance guidance, and voice interaction (11). The maintenance feedback layer includes three functional modules: maintenance result feedback (12), collection of maintenance feedback data (13), and updating of the diagnostic rule base (14).

[0048] To further optimize the above technical solution, this invention first performs vehicle startup 1 in the data acquisition layer (vehicle edge layer). After vehicle startup, data is collected by a multi-sensor data acquisition device 2, including image and video data collected by a camera 2-1, point cloud data collected by a lidar 2-2 to construct a 3D map, mid-to-long-range data information collected by a millimeter-wave radar 2-3, near-range related data collected by an ultrasonic radar 2-4, and real-time vehicle positioning information and predicted path planning navigation information collected by a GUSS / IMU 2-5 inertial navigation system. After the vehicle environmental perception data is collected, data preprocessing is performed. First, time synchronization 3-1 is performed. In the detection data, since there are more or less time differences and time drifts in the detection data points of various sensors, it is necessary to solve the problems of time drift, zero-point drift, and reference calibration. In this module, multi-sensor fusion is used to remove time drift, zero-point drift, and reference calibration problems by using wavelet analysis, so that each perception module has the same time reference.

[0049] To further optimize the above technical solution, coordinate system transformation (function 3-2) is also required. Inconsistent positioning coordinates among the camera (2-1), lidar (2-2), millimeter-wave radar (2-3), ultrasonic radar (2-4), and GUSS / IMU (2-5) can lead to distortion and deformation of the constructed object and the road. Based on incorrect perception data and incorrect vehicle positioning, decision-making errors will occur, resulting in a decrease in overall vehicle comfort and a reduction in safety.

[0050] To further optimize the above technical solutions, in the basic feature extraction 3-3, the utilization rate of effective parameters such as the SSIM structural similarity of the image and the signal-to-noise ratio of the radar can directly affect the execution of the perception system, and specific parameters need to be extracted according to different characteristics. Due to the particularly large amount of high-precision map data collected, functional modules such as Gini coefficient data compression 3-4 are also needed to enable rapid data flow.

[0051] To further optimize the above technical solution, in the cloud analysis layer, hardware fault detection 4 is mainly responsible for hardware inspection of the perception system, including hardware and software Ping value matching, CAN communication, wiring harness connection faults, hardware identification and other faults. Within the rules of rule engine 4-1, based on the database, intelligent judgment and identification are performed through intelligent matching and filtering to achieve hardware judgment.

[0052] In previous troubleshooting processes, hardware diagnostics were typically performed using diagnostic tools to read fault codes. The direction of these fault codes was then used to determine the next steps in testing and troubleshooting, with real-time fault codes provided as feedback. In hardware fault code detection 4-2, by importing all fault codes related to the sensing system from the database and comprehensively assessing the diagnostic results through multi-faceted diagnostics, the system can intelligently pinpoint the more precise fault location, shortening troubleshooting time.

[0053] The response time of sensors in a sensing system greatly affects data fitting and data iteration time. In section 4-3, a response time of less than 100ms is considered good. When the iteration time is greater than 100ms, the system automatically judges that the sensor data is invalid or the data is delayed, and it is necessary to reload the relevant data from other sensors to fit all the required special data.

[0054] To further optimize the above technical solution, when hardware fault detection 4, rule engine 4-1, and hardware fault code detection 4-2 determine through rules that a fault exists, it is considered that the intelligent connected vehicle perception system cannot perform its perception function. In this case, it will immediately enter degraded mode 5-1, safely stop the vehicle, and ensure the safety of the driver and passengers.

[0055] To further optimize the above technical solution, if no fault risk points are detected, the process proceeds to extract feature parameters 5-2, such as the SSIM structural similarity of the image and the signal-to-noise ratio of the radar, which directly affect the execution of the perception system. A full-stack early warning and data closed-loop system is implemented, with early warning levels determined based on data analysis. Early warning levels are: Level 1 (Alert): Performance degradation <10%, observation recommended; Level 2 (Warning): Performance degradation 10-30%, maintenance scheduled; Level 3 (Emergency): Performance degradation >30%, vehicle speed limit triggered. Different vehicle functions are implemented according to different levels.

[0056] Because each sensor performs a different function, multi-sensor fusion technology is used here. Five different types of sensors, namely camera 2-1, lidar 2-2, millimeter-wave radar 2-3, ultrasonic radar 2-4, and GUSS / IMU 2-5, serve as data acquisition inputs. The main items and characteristic parameters detected by each sensor are the primary basis. When a sensor fails or the data is inaccurate, it is necessary to use the characteristic parameters of other sensors to fit the parameters to compensate for the missing characteristic values ​​of the failed sensor and ensure safe driving.

[0057] To further optimize the above technical solution, the cloud layer mainly receives feature vectors and feature points uploaded from the data acquisition edge layer, and runs a multimodal diagnostic model group, including anomaly detection model, health prediction model, and collaborative verification module.

[0058] The anomaly detection model, based on the Temporal Convolutional Network (TCN), identifies sensor data drift and can combine data processing to analyze data drift problems. X(n) represents the nth standardized reconstruction time sample, as shown in the formula.

[0059] X(n)=(x1 n …,x t n ,…,x T n )∈R T×M

[0060] In the formula: T is the time window length; M is the number of sensors; x t n The sensor time series of the nth standardized sample at the tth sampling time can be represented by the vector of the sensor time series of the nth destandardized reconstructed sample at the tth sampling time.

[0061] x t n =(x t,1 n ,x t,2 n ,…,x t,Mn )T∈R M

[0062] After data acquisition, it is transmitted to the cloud via cloud data transfer (6). This process must ensure both the authenticity and timeliness of the data flow. In the multi-level data analysis (7) in the cloud, lifespan prediction (7-1), health assessment (7-2), and cross-sensor verification (7-3) are performed. In the lifespan prediction (7-1) module, sensor lifespan is comprehensively judged based on the sensor's response time, service life, operating time, communication speed, and data quality.

[0063] To further optimize the above technical solutions, in the health assessment (7-2), a health prediction model is constructed, and a health function is built. Based on point cloud density, data clarity, data validity, and data noise distribution, combined with the predicted values ​​from the lifespan prediction module (7-1) of lidar sensors, cameras, millimeter-wave radar, ultrasonic radar, and GNSS / IMU inertial navigation, a comprehensive judgment is given through multi-sensor fusion. Cross-sensor verification (7-3) is for collaborative verification modules. It detects latent faults through cross-sensor association rule bases (such as camera-radar ranging consistency thresholds), addressing the pain points of software mismatch and hardware latent sub-state faults.

[0064] To further optimize the above technical solution, based on the results of multi-level cloud data analysis 7, data analysis is conducted on vehicle attitude, vehicle speed, vehicle heading angle, vehicle chassis parameters, power system parameters, and overall vehicle parameters to ultimately achieve dynamic decision-making 8.

[0065] After dynamic decision-making step 8 is completed, fault level judgment 9 is performed, and different levels correspond to different responses. For general faults with functional degradation, performance degradation needs to be recorded 9-1 for feedback to the cloud big data end, forming a closed feedback data structure.

[0066] To further optimize the above technical solution, when the diagnosis indicates that repair or maintenance is required, a scheduled repair reminder 9-2 will be activated to remind customers to enter the maintenance phase on time and ensure customer safety.

[0067] To further optimize the above technical solution, when the prompt function is limited or the fault is serious, the vehicle speed limit control 9-3 and other function modules will be entered and used in a downgraded manner, and the vehicle will need to be stopped if necessary.

[0068] To further optimize the above technical solution, the decision execution layer includes two functional modules: maintenance guidance generation 10 and AR maintenance guidance, and voice interaction 11.

[0069] First, in the maintenance guidance generation 10, a clear maintenance guidance document is given based on real-time feedback data and vehicle parameters to determine the specific parts and software to be repaired or maintained; then, AR maintenance guidance and voice interaction 11 are introduced to help customers understand the fault point repair and how to provide feedback to maintenance professionals.

[0070] The maintenance feedback layer includes three functional modules: maintenance result feedback (12), maintenance feedback data collection (13), and diagnostic rule base update (14).

[0071] To further optimize the above technical solutions, in the maintenance result feedback 12, the feedback mechanism will statistically send the information to the cloud big data to establish a complete database. The database is updated in real time. The longer the time, the larger the data becomes, and the more accurate the subsequent fault detection will be. In the maintenance feedback data collection 13 module, the data will be effectively filtered and updated to the updated diagnostic rule base 14 in a timely manner. The rule base will then feed back to the cloud for multi-level data analysis.

[0072] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0073] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A diagnostic system for a sensing module of an intelligent connected vehicle, characterized in that, include: The data exchange consists of a data acquisition layer, a cloud analysis layer, a decision execution layer, and a maintenance feedback layer. The data acquisition layer, deployed in the vehicle-mounted edge computing unit, is used to acquire raw data from cameras, lidar, millimeter-wave radar, ultrasonic sensors, and GNSS / IMU in real time, and to perform time synchronization, coordinate system transformation, feature extraction, and data compression. The cloud-based analysis layer receives feature vectors uploaded from the edge layer and runs a multimodal diagnostic model group, including an anomaly detection model, a health prediction model, and a collaborative verification module. The decision-making and execution layer generates maintenance guidance strategies based on fault levels, supporting AR visualization and voice interaction; The maintenance feedback layer collects maintenance result data and dynamically updates the diagnostic rule base.

2. The intelligent connected vehicle perception module diagnostic system according to claim 1, characterized in that, The data acquisition layer includes vehicle startup, multi-sensor data acquisition, and data preprocessing.

3. The intelligent connected vehicle perception module diagnostic system according to claim 2, characterized in that, The multi-sensor data acquisition device includes a camera, lidar, millimeter-wave radar, ultrasonic radar, and GUSS / IMU; the multi-sensor data acquisition device includes time synchronization, coordinate transformation, feature extraction, and data compression.

4. The intelligent connected vehicle perception module diagnostic system according to claim 1, characterized in that, The cloud-based analysis layer includes hardware fault detection, degradation mode, safe parking, feature parameter extraction, cloud data transmission, multi-level cloud data analysis, dynamic decision-making, and fault level judgment.

5. The intelligent connected vehicle perception module diagnostic system according to claim 4, characterized in that, The hardware fault detection includes a rule engine, hardware fault code detection, and response time.

6. The intelligent connected vehicle perception module diagnostic system according to claim 4, characterized in that, The cloud-based multi-level data analysis includes lifespan prediction, health assessment, and cross-sensor verification.

7. The intelligent connected vehicle perception module diagnostic system according to claim 4, characterized in that, The fault level determination includes recording performance degradation, scheduling maintenance reminders, and vehicle speed limit control.

8. The intelligent connected vehicle perception module diagnostic system according to claim 1, characterized in that, The decision execution layer includes maintenance guidance generation, AR maintenance guidance, and voice interaction.

9. The intelligent connected vehicle perception module diagnostic system according to claim 1, characterized in that, The maintenance feedback layer includes maintenance result feedback, collection of maintenance feedback data, and updating of the diagnostic rule base.

10. A diagnostic method for the perception module of an intelligent connected vehicle, characterized in that, The automatic vehicle diagnostic system achieved using the intelligent connected vehicle perception module diagnostic system described in claims 1-9 includes the following steps: Step S1: First, start the vehicle. After the vehicle starts, data is collected through the multi-sensor data acquisition unit. Step S2: After the vehicle environmental perception data is collected, data preprocessing is performed. First, time synchronization is performed to solve problems such as time drift, zero drift, and benchmark calibration. Then, coordinate system transformation is performed. After completion, basic feature extraction is performed. Since the amount of high-precision map data collected is particularly large, Gini data compression is also required to enable data to flow quickly. Step S3: After data preprocessing is completed, hardware fault detection is performed. This mainly involves hardware checks of the sensing system, including hardware and software ping value matching, CAN communication, wiring harness connection faults, hardware identification, and other fault detection. Hardware fault code detection is also performed to intelligently locate the fault point more accurately, shortening the troubleshooting time. After the troubleshooting time is shortened, the sensing system sensors are activated. The response time of the sensing system sensors is specified in 4-3 as less than 100ms to be considered good. When the iteration time is greater than 100ms, the system automatically judges that the sensor data is invalid or the data is delayed, and it is necessary to reload the relevant data of other sensors to fit all the special data required. Step S4: When a hardware fault is detected, the system immediately enters degrade mode, safely stops the vehicle, and ensures the safety of the driver and passengers. If no fault risk is detected, the system extracts feature parameters, performs full-stack early warning and data closure, and implements warning classification based on data judgment. Different vehicle functions are implemented according to different levels. Since the functions performed by each sensor are different, multi-sensor fusion technology is used here. Five different types of sensors are used as data acquisition inputs, and the main items and feature parameters detected by each sensor are the main basis. When a sensor fails or the data is inaccurate, it is necessary to use the feature parameters of other sensors to fit parameters to compensate for the missing feature values ​​of the failed sensor and ensure safe driving. Step S5: After the feature parameters are acquired, the data is transmitted to the cloud for multi-level data analysis, including lifespan prediction, health assessment, and cross-sensor verification. Step S6: Based on the results of multi-level data analysis in the cloud, the vehicle attitude, vehicle speed, vehicle heading angle, vehicle chassis parameters, power system parameters and vehicle parameters are analyzed to ultimately achieve dynamic decision-making; Step S7: After dynamic decision-making is completed, the fault level is determined, and different levels correspond to different responses. For general faults with functional degradation, performance degradation needs to be recorded and fed back to the cloud big data platform to form a closed feedback data structure. When the diagnostic prompts that repair or maintenance is needed, a repair appointment reminder is initiated to remind the customer to proceed with maintenance on time, ensuring customer safety. When the prompt indicates limited functionality or a more serious fault, the vehicle speed limit control module is activated, and its operation is downgraded; if necessary, the vehicle must be stopped. Step S8: After determining the fault level, two functions are used: maintenance guidance generation and AR maintenance guidance, and voice interaction. Based on real-time feedback data and vehicle parameters, a clear maintenance guidance document is provided to determine the specific parts and software to be repaired or maintained. Then, AR maintenance guidance and voice interaction are introduced to help customers understand the fault point repair and how to provide feedback to professional maintenance personnel. Step S9: After completion, feedback on the repair results is provided. The feedback mechanism will statistically analyze the information and send it to the cloud big data platform to establish a comprehensive database. In the module for collecting repair feedback data, the data will be effectively filtered and updated in a timely manner to the diagnostic rule base. The rule base will then feed the data back to the cloud for multi-level data analysis.

Images