Intelligent connected vehicle automatic driving control method and control system thereof

Abstract

This invention discloses an autonomous driving control method and control system for intelligent connected vehicles. The control method includes: collecting multi-source onboard data, extracting continuity features, consistency features, and temporal trend features from the multi-source data, and identifying fault types by combining them with preset joint judgment rules; constructing a two-dimensional hierarchical system of fault severity index and scenario risk index, calculating a hierarchical score based on the fault type, and determining a downgrade level in the two-dimensional hierarchical system based on the hierarchical score; setting dynamic constraints according to the downgrade level, searching for safe parking domains in real time, and generating parking paths that meet the dynamic constraints and obstacle avoidance requirements; controlling the vehicle to travel along the parking path to the safe parking domain, and performing safe parking and warning operations. The technical solution of this application can effectively control the vehicle and improve safety when a fault occurs.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and more specifically to an autonomous driving control method and control system for intelligent connected vehicles. Background Technology

[0002] With the rapid development of intelligent connected vehicle technology, the safety of autonomous driving systems has become a core issue for industrialization. However, existing autonomous driving control methods suffer from low accuracy in fault detection, are susceptible to external interference leading to misjudgments and missed detections, and cannot provide reliable support for vehicle control. In subsequent vehicle control and safe parking processes, issues such as unstable vehicle operation and improper parking can easily arise, making it difficult to ensure the safety of the vehicle and surrounding road users. In short, existing control schemes are inadequate for safe emergency control when faults occur. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an intelligent connected vehicle autonomous driving control method that can effectively control the vehicle and improve safety when a malfunction occurs.

[0004] This application provides an autonomous driving control method for intelligent connected vehicles, the control method comprising:

[0005] Collect multi-source data from the vehicle, extract the continuity features, consistency features and time-series trend features of the data from the multi-source data, and identify the fault type by combining the preset joint judgment rules;

[0006] A two-dimensional classification system of fault severity index and scenario risk index is constructed. A classification score is obtained by weighted calculation according to the fault type. The downgrade level is determined in the two-dimensional classification system based on the classification score.

[0007] Based on the degradation level, dynamic constraints are set, safe parking areas are searched in real time, and parking paths that meet the dynamic constraints and obstacle avoidance requirements are generated.

[0008] Control the vehicle to drive along the parking path to the safe parking area, and perform safe parking and warning operations.

[0009] In one aspect, the vehicle-mounted multi-source data includes perception data, execution data, and communication data. The perception data includes observation data from various vehicle-mounted sensors, the execution data includes operating status data of vehicle-mounted actuators and vehicle operating parameters, and the communication data includes vehicle-to-everything (V2X) communication data and operating status data of the vehicle-mounted computing platform.

[0010] The steps for collecting multi-source data from vehicles include:

[0011] Determine the collection range of vehicle-mounted multi-source data, and synchronously collect vehicle-mounted multi-source data based on the collection range;

[0012] The collected data are categorized by data type.

[0013] In one aspect, the steps following the acquisition of multi-source vehicle data include:

[0014] The multi-source data is preprocessed, the filtering threshold is adjusted based on the data type, invalid data is removed according to the filtering threshold, and the data format is standardized.

[0015] In one aspect, the continuity feature is used to represent the stability of data updates, the consistency feature is used to represent the matching degree between multi-source data instructions and feedback, and the time-series trend feature is used to represent the cumulative characteristics of abnormal data.

[0016] The steps for extracting the continuity features, consistency features, and time-series trend features of the data from the multi-source data include:

[0017] The continuity feature, consistency feature, and time-series trend feature of the data are extracted separately to obtain the continuous feature extraction value, consistency feature extraction value, and time-series trend feature extraction value.

[0018] Initial extraction weights are assigned to the continuity feature, the consistency feature, and the time-series trend feature, wherein the initial weights are preset according to the multi-source data type;

[0019] The initial extraction weights are dynamically calibrated based on the real-time driving scenario of the vehicle.

[0020] The calibrated initial extraction weights are weighted and calculated with the corresponding feature extraction values ​​to obtain the fused feature comprehensive value.

[0021] In one aspect, the steps for identifying fault types by combining preset joint determination rules include:

[0022] The system determines the anomalies based on a preset threshold and duration, wherein the preset threshold is updated in real time according to the vehicle's driving status and environmental changes.

[0023] When at least two features are abnormal at the same time, it is determined to be a real fault and the fault type is identified;

[0024] When a single feature is abnormal, it is judged as noise interference and fault type identification is not triggered.

[0025] In one aspect, the steps of constructing a two-dimensional classification system of fault severity index and scenario risk index, and obtaining a classification score by weighting the scores according to the fault type, include:

[0026] A two-dimensional classification system is constructed, which includes a fault severity index and a scenario risk index. The fault severity index is used to represent the severity of the fault itself, and the scenario risk index is used to represent the risk level of the vehicle's current driving scenario.

[0027] The fault severity index is assigned a first preset weight, and the scenario risk index is assigned a second preset weight, wherein the first preset weight is adjusted in real time according to the fault type, and the second preset weight is adjusted in real time according to the current scenario risk level.

[0028] Based on the fault type, determine the corresponding fault severity index value, and combine it with the real-time driving scenario of the vehicle to determine the scenario risk index value.

[0029] The severity of the fault index and its corresponding first preset weight, as well as the scenario risk index and its corresponding second preset weight, are weighted and calculated to obtain a graded score.

[0030] In one aspect, the step of determining a downgrade level in the two-dimensional grading system based on the grading score includes:

[0031] The graded score is divided into at least three intervals, each interval corresponding to a downgrade level, and the downgrade level is positively correlated with the score interval;

[0032] The graded score is obtained, and the graded score is substituted into a preset score range to obtain the preliminary result of the corresponding downgrade level.

[0033] By combining the assigned values ​​of the fault severity index and the scenario risk index, the preliminary result of the downgrade level is verified, and the verified downgrade level is output.

[0034] In one aspect, the dynamic constraints include deceleration constraints, steering constraints, and control quantity smoothing constraints;

[0035] The steps for setting dynamic constraints based on the degradation level include:

[0036] Establish the correspondence between the degradation levels and constraint parameters;

[0037] Based on the degradation level, a corresponding constraint parameter is matched, wherein the degradation level is positively correlated with the strictness of the constraint parameter;

[0038] The dynamic constraints for controlling the vehicle are formed based on the aforementioned constraint parameters;

[0039] Based on the verified downgrade level, the corresponding dynamic constraints are updated to match the constraint parameters with the current downgrade level and vehicle driving state.

[0040] In one aspect, the step of setting dynamic constraints based on the degradation level further includes:

[0041] Based on the vehicle's current downgrade level and driving status, the cutoff frequency of the low-pass filter is determined, and transient fluctuations in the vehicle control quantities are filtered based on the cutoff frequency. The control quantities include braking pressure, steering angle, and motor torque.

[0042] Based on the constraint parameters corresponding to the downgrade level, a dynamic limiting range is set, wherein the dynamic limiting range includes the maximum fluctuation range and the rate of change limit of the control quantity;

[0043] The control quantity filtered by the low-pass filter is substituted into the dynamic limiting range for verification. If the control quantity exceeds the dynamic limiting range, the control quantity is adjusted to the dynamic limiting range.

[0044] Furthermore, to address the aforementioned issues, this application also provides an intelligent connected vehicle autonomous driving control system, the control system comprising:

[0045] The identification module is used to collect multi-source data from the vehicle, extract the continuity features, consistency features and time-series trend features of the data from the multi-source data, and identify the fault type by combining the preset joint judgment rules.

[0046] The calculation module is used to construct a two-dimensional classification system of fault severity index and scenario risk index, perform weighted calculation according to the fault type to obtain a classification score, and determine the downgrade level in the two-dimensional classification system based on the classification score;

[0047] The search module is used to set dynamic constraints according to the degradation level, search for safe parking areas in real time, and generate parking paths that meet the dynamic constraints and obstacle avoidance requirements.

[0048] The control module is used to control the vehicle to travel along the parking path to the safe parking area and to perform safe parking and warning operations.

[0049] The beneficial effects of this invention are as follows: it can effectively control the vehicle and improve the safety of the autonomous driving system when a fault occurs; in the early stage of a fault, it can identify the fault situation and reduce control failures caused by fault identification errors; by determining the degradation level, it can match the appropriate control mode according to the actual fault situation and driving scenario, reducing the safety risks caused by over-control or under-control; through vehicle control and safe parking operations, it can ensure smooth vehicle operation and standardized parking, effectively reducing safety hazards such as vehicle jerking, skidding, and rear-end collisions, and protecting the safety of the vehicle itself and surrounding traffic participants; it can achieve effective management and control after a fault occurs, reduce the incidence of safety accidents caused by faults, and improve the safety and reliability of the autonomous driving system of intelligent connected vehicles. Attached Figure Description

[0050] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0051] Appendix Figure 1 This is a flowchart illustrating the intelligent connected vehicle autonomous driving control method of this application;

[0052] Appendix Figure 2 This is a functional structure diagram of the intelligent connected vehicle autonomous driving control system of this application. Detailed Implementation

[0053] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0054] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0055] like Figure 1 As shown, this application provides an autonomous driving control method for intelligent connected vehicles. Intelligent connected vehicles are equipped with onboard perception, decision-making, and execution intelligent devices, integrating modern communication and network technologies to achieve comprehensive information interaction and data sharing between vehicles, between vehicles and road infrastructure, between vehicles and people, and between vehicles and cloud-based backends. They possess autonomous driving capabilities such as environmental perception, intelligent decision-making, and automatic control. Through the synergistic integration of intelligent driving and connectivity technologies, this new generation of vehicles can improve driving safety, traffic efficiency, and intelligent control levels, and is also an important carrier for the combination of autonomous driving technology and vehicle-to-everything (V2X) technology. The control method includes:

[0056] Step S10: Collect multi-source vehicle data. Extract continuity features, consistency features, and temporal trend features from the multi-source data, and identify the fault type by combining them with preset joint judgment rules. During the autonomous driving process, multi-source vehicle data is collected, including various operational data related to vehicle perception, execution, and communication. After data collection, continuous features reflecting data update stability, consistency features reflecting the matching degree between multi-source data commands and feedback, and temporal trend features reflecting the accumulation characteristics of abnormal data are extracted from the multi-source data. After extraction, the vehicle fault type is identified by combining the extracted features with preset joint judgment rules. The joint judgment rules can effectively distinguish between real faults and feature anomalies caused by noise interference, ensuring the accuracy of fault type identification.

[0057] Commonly used vehicle perception sensors include LiDAR, millimeter-wave radar, vision cameras, ultrasonic radar, inertial measurement units (IMUs), tire pressure sensors, and vehicle speed sensors. LiDAR is often installed on the roof or in the center of the front bumper. Millimeter-wave radar is placed in the front grille and at the four corners of the front and rear bumpers. Vision cameras are divided into forward-looking and surround-view types. Forward-looking cameras are installed behind the rearview mirrors on the windshield, while surround-view cameras are placed below the front and rear bumpers and the left and right rearview mirrors. Ultrasonic radar is embedded inside the front and rear bumpers. Inertial measurement units and Beidou positioning modules are often integrated inside the center console or in a stable position on the chassis. Tire pressure sensors are built into each tire, and vehicle speed sensors are installed at the wheel hubs. Execution-related status detection sensors include steering angle sensors and brake pressure sensors. Steering angle sensors are mounted on the steering column, while brake pressure sensors are located in the master cylinder or brake lines. Communication-related vehicle networking transceiver modules and in-vehicle Ethernet communication detection components are often installed inside the B-pillar or next to the electrical box on the center console.

[0058] Step S20: Construct a two-dimensional classification system of fault severity index and scenario risk index. A classification score is obtained by weighted calculation based on the fault type. The downgrade level is determined within the two-dimensional classification system based on the classification score. The fault severity index characterizes the severity of the fault itself, while the scenario risk index characterizes the risk level of the vehicle's current driving scenario. Based on the identified fault type, the fault severity index and scenario risk index are assigned values ​​and corresponding preset weights. A classification score is obtained through weighted calculation, and then combined with the classification score into the two-dimensional classification system to match and determine the corresponding vehicle autonomous driving downgrade level, making the determination of the downgrade level more consistent with the actual operating state of the vehicle.

[0059] Step S30: Set dynamic constraints according to the downgrade level, search for safe parking areas in real time, and generate parking paths that meet the dynamic constraints and obstacle avoidance requirements. The dynamic constraints include control requirements related to vehicle dynamics, such as deceleration constraints, steering constraints, and control quantity smoothing constraints during vehicle driving. After the dynamic constraints are set, the driving environment around the vehicle is detected and analyzed in real time to search for safe parking areas that meet the safe parking requirements. At the same time, combined with the set dynamic constraints and obstacle avoidance requirements of vehicle driving, a corresponding parking path is generated. The parking path not only meets the control constraints of vehicle dynamics, but also effectively avoids various obstacles during driving, ensuring the safety of the vehicle in the process of driving to the safe parking area.

[0060] Step S40: Control the vehicle to travel along the parking path to the safe parking area and perform safe parking and warning operations. Based on the generated parking path, control the vehicle's driving status, guiding it smoothly along the path to the designated safe parking area. Once the vehicle reaches the safe parking area, perform a safe parking operation to maintain a stable parking position. Simultaneously, activate the corresponding warning operation to transmit vehicle malfunction information to surrounding traffic participants and the relevant traffic monitoring platform. This allows surrounding traffic participants to promptly notice the vehicle's status and enables the monitoring platform to grasp the actual situation of vehicle malfunction emergency handling, ensuring the safety of the vehicle and surrounding traffic participants.

[0061] In this embodiment, the vehicle can be effectively controlled and the safety of the autonomous driving system can be improved when a fault occurs. In the early stage of a fault, the fault situation can be identified to reduce control failures caused by fault identification errors. By determining the degradation level, the control mode can be matched and adapted according to the actual fault situation and driving scenario to reduce the safety risks caused by over-control or under-control. Through vehicle control and safe parking operations, the vehicle can be ensured to run smoothly and park properly, effectively reducing safety hazards such as vehicle jerking, skidding, and rear-end collisions, and protecting the safety of the vehicle itself and surrounding traffic participants. Effective management and control after a fault occurs can be achieved, reducing the incidence of safety accidents caused by faults and improving the safety and reliability of the autonomous driving system of intelligent connected vehicles.

[0062] In one aspect, vehicle-mounted multi-source data includes perception data, execution data, and communication data. Perception data includes observation data from various vehicle-mounted sensors, execution data includes the working status data of vehicle-mounted actuators and vehicle operating parameters, and communication data includes vehicle-to-everything (V2X) communication data and the operating status data of the vehicle-mounted computing platform.

[0063] The steps for collecting multi-source data from vehicles include:

[0064] Step S110: Determine the collection range of vehicle-mounted multi-source data, and collect vehicle-mounted multi-source data synchronously based on the collection range. The collection range is defined according to the needs of autonomous driving fault detection to ensure that the collected data can support subsequent feature extraction and fault identification. After determining the collection range, extract all vehicle-mounted multi-source data within the range in a synchronous collection manner to ensure the consistency of various types of data in the time dimension and reduce the reduction of data effectiveness due to collection time difference.

[0065] Step S111 involves classifying the collected data by data type. All collected data is classified according to the criteria of perception data, execution data, and communication data. This process ensures the orderly organization of the collected data, enabling targeted feature extraction for different data types. It also reduces interference from mixed data types in subsequent data preprocessing and feature extraction, improving the overall efficiency and accuracy of data processing.

[0066] In one embodiment of this application, after the step of collecting vehicle-mounted multi-source data, the following steps are included:

[0067] Step S120 involves preprocessing the multi-source data, adjusting the filtering threshold based on data type, removing invalid data according to the filtering threshold, and standardizing the data format. After classifying the vehicle-mounted multi-source data, preprocessing is performed on the classified multi-source data. Based on the different data type characteristics of sensing, execution, and communication types, the filtering thresholds used for data filtering are adjusted accordingly. Then, using the adjusted filtering thresholds as the judgment standard, invalid data without actual data value is identified and removed, reducing the interference of invalid data on subsequent feature extraction and fault identification. Simultaneously, the format of the filtered valid data is standardized, ensuring that multi-source data of different types and sources form a unified format standard, such as CAN bus format, Ethernet message format, or JSON format.

[0068] Taking the processing of ultrasonic radar data for vehicle-mounted sensing as an example, in view of the characteristic that ultrasonic radar is susceptible to environmental clutter interference and generates invalid ranging data, an appropriate distance filtering threshold is adjusted for it. For example, the effective ranging filtering threshold is set to 0.1 meters to 5 meters. Using the adjusted filtering threshold as the judgment standard, the multi-source ranging data collected by ultrasonic radar is identified one by one. False ranging data below 0 meters caused by abnormal sensor signal reflection, clutter ranging data above 5 meters that has no actual detection significance, and invalid data with jumps caused by temporary electromagnetic interference are identified and directly removed from the multi-source data collected by the radar, retaining only the effective ranging data in the range of 0.1 meters to 5 meters.

[0069] In one embodiment of this application, the continuity feature is used to represent the data update stability, the consistency feature is used to represent the matching degree between multi-source data instructions and feedback, and the time-series trend feature is used to represent the cumulative characteristics of abnormal data.

[0070] The steps for extracting continuity features, consistency features, and time-series trend features from multi-source data include:

[0071] Step S130 involves extracting the continuity, consistency, and time-series trend features of the data, respectively, to obtain the extracted values ​​for the continuity, consistency, and time-series trends. By using corresponding extraction methods, values ​​that quantify each feature are obtained, ensuring that the characteristics of each feature are reflected through specific numerical values.

[0072] For continuous feature extraction, the update time interval deviation method can be used. This method is suitable for real-time data acquisition from sensors, actuators, and other devices with a fixed update frequency. Generally, a preset update frequency is first determined, and the actual update time intervals of several consecutive frames are extracted. The continuous feature extraction value is obtained by calculating the average or maximum deviation between the actual interval and the preset interval. The smaller the continuous feature extraction value, the stronger the stability of the data in the update time dimension, and the better the continuity of data updates. Alternatively, the unit time frame loss rate statistical method can be used. For various types of continuously acquired time-series data, the difference between the actual number of acquired frames and the theoretically required number of acquired frames within a fixed time window is statistically analyzed. The frame loss rate or its complement compliance rate is then calculated as the continuous feature extraction value. The lower the frame loss rate and the higher the compliance rate, the fewer frames are lost during data acquisition, and the better the data continuity. Another method is the data transmission integrity rate method, which is specifically used for communication data such as vehicle network interactive data and vehicle computing platform instruction data. It uses a single fixed data packet as the statistical unit and calculates the data transmission integrity rate by the ratio of the number of valid transmitted bytes in the packet to the total number of bytes that should be transmitted. The integrity rate is used as the continuity feature extraction value. The level of the integrity rate directly reflects the integrity and continuity of communication data during the transmission process. The higher the integrity rate, the better the transmission continuity.

[0073] For consistency feature extraction methods, the difference statistical quantification method can be used. This method is suitable for multi-source sensing data or actuator command-feedback data in the same scenario. It extracts two sets of data to be matched at the same timestamp, and obtains the consistency feature extraction value by calculating the average absolute deviation or root mean square difference between the two sets of data. The smaller the value, the smaller the numerical deviation between the two sets of data, and the higher the matching degree and consistency between the data. Alternatively, the data correlation analysis method can be used. For several consecutive sets of multi-source homogeneous data or command-feedback data, the correlation operation is performed on the two sets of data to obtain the Pearson correlation coefficient, and this coefficient is used as the consistency feature extraction value. The closer the coefficient value is to 1, the higher the linear correlation between the two sets of data, and the stronger the consistency between the data. Another method is the command-feedback consistency method, used for discrete command data such as motor torque output and parking brake start / stop. It counts the proportion of times the actual feedback state after the actuator receives a command within a fixed time period is consistent with the command requirement, and uses this proportion as the consistency feature extraction value. The higher the proportion, the higher the matching degree between the actuator's command reception and actual execution, and the better the consistency between command and feedback.

[0074] For extracting time-series trend features, the time-series deviation summation method can be used. This method is applicable to all continuous time-series data collected by vehicles. First, the normal threshold or mean of this type of data is determined through vehicle operation calibration. The deviation value of each frame of data in several consecutive frames from the normal threshold is extracted. All deviation values ​​are summed to obtain the extracted time-series trend feature value. The magnitude of this summation value directly reflects the degree of accumulation of abnormal data. The smaller the value, the less the data deviates from the normal state and the more stable the time-series trend. Alternatively, the sliding window trend slope method can be used for analyzing abnormal trend changes in continuous time-series data. First, a sliding time window with a fixed duration or a fixed number of frames is set. Data windows are captured sequentially, and the slope of the data deviation value within each window is calculated. This slope is used as the extracted time-series trend feature value. The closer the slope value is to 0, the less obvious the change trend of data deviation and the less significant the accumulation of abnormal data. The larger the absolute value of the slope, the more obvious the change trend of data deviation and the faster the rate of abnormal accumulation. Another method is the cumulative percentage of abnormal data. For various types of time-series data, the cumulative number of frames or cumulative duration of abnormal data that exceed the normal threshold range within a fixed time window is counted. The proportion of this to the total number of frames or total duration of the window is calculated as the time-series trend feature extraction value. The lower the proportion, the less the amount of abnormal data accumulated within the window and the more stable the time-series trend of the data. Conversely, the higher the proportion, the more serious the accumulation of abnormal data.

[0075] Step S131: Assign initial extraction weights to continuity features, consistency features, and time-series trend features. The initial weights are pre-set based on the multi-source data types. The initial extraction weights are pre-set based on different multi-source data types such as perception, execution, and communication. The initial extraction weights are set in combination with the feature attributes of different data types and their impact on fault identification, so that the initial extraction weight allocation matches the actual characteristics of the data types.

[0076] Step S132: Dynamically calibrate the initial extraction weights in conjunction with the real-time driving scenario of the vehicle; perform dynamic calibration operation on the set initial extraction weights, and adjust the extraction weight values ​​corresponding to each feature according to the actual situation such as risk changes and road condition differences in the vehicle driving scenario, so that the extraction weights can adapt to the real-time driving status of the vehicle.

[0077] Step S133: The calibrated initial extraction weights are weighted and calculated with the corresponding feature extraction values ​​to obtain the fused feature comprehensive value. The calibrated feature extraction weights are weighted and calculated with the corresponding continuous feature extraction values, consistency feature extraction values, and time-series trend feature extraction values. The quantified data of the three types of features are fused through weighted calculation to obtain the fused feature comprehensive value that can comprehensively represent the multi-dimensional features of the data.

[0078] F = wc × C + wy × Y + ws × S, where F is the comprehensive value of the fused features; wc is the extraction weight of continuous features after dynamic calibration, with a value range of 0-1; C is the extracted value of continuous features; wy is the extraction weight of consistency features after dynamic calibration, with a value range of 0-1; Y is the extracted value of consistency features; ws is the extraction weight of time-series trend features after dynamic calibration, with a value range of 0-1; wc + wy + ws = 1.

[0079] In one embodiment of this application, the step of identifying fault types by combining preset joint determination rules includes:

[0080] Step S140: Determine feature anomalies based on preset thresholds and duration. The preset thresholds are updated in real time according to vehicle driving status and environmental changes. The preset thresholds are not fixed values ​​and are updated in real time according to the current driving status of the vehicle and changes in the external environment, so that the preset threshold judgment criteria can be adapted to the actual operating conditions of the vehicle. Through the dual judgment of preset thresholds and duration, accurate identification of whether features are abnormal can be achieved.

[0081] Furthermore, when conducting feature anomaly determination, the preset threshold of the fusion feature comprehensive value can be determined and updated in real time based on the vehicle's driving status and environmental changes, forming a clear numerical judgment boundary between normal and abnormal features. Subsequently, the fusion feature comprehensive value calculated in real time during vehicle operation is continuously monitored. On the one hand, it is determined whether the fusion feature comprehensive value exceeds the preset threshold range. On the other hand, the actual duration of the fusion feature comprehensive value being outside the preset threshold range is recorded simultaneously. Then, a comprehensive judgment is made in combination with the preset abnormal duration threshold. When the fusion feature comprehensive value exceeds the preset threshold and the duration of the abnormal value state reaches or exceeds the preset abnormal duration threshold, the corresponding feature is determined to be abnormal. If the fusion feature comprehensive value does not exceed the preset threshold, or only briefly exceeds the threshold but the duration does not reach the abnormal duration threshold, it is not determined to be a feature anomaly.

[0082] Step S141: When at least two features are abnormal at the same time, it is determined to be a real fault and the fault type is identified. When at least two features among the continuity feature, consistency feature, and time-series trend feature are determined to be abnormal at the same time, the vehicle is identified to have a real fault. At this time, the specific fault type of the vehicle is identified according to the combination type of abnormal features and the corresponding feature performance.

[0083] In step S142, if a single feature is abnormal, it is determined to be noise interference, and fault type identification is not triggered. If an abnormality is detected in a single feature among continuous features, consistent features, and time-series trend features, it is determined that the abnormality is caused by external noise interference and is not an actual vehicle fault. In this case, the vehicle fault type identification process is not triggered, reducing misjudgments caused by noise interference, reducing unnecessary vehicle control operations, and ensuring the stability of autonomous driving operation.

[0084] In one embodiment of this application, the step of constructing a two-dimensional classification system of fault severity index and scenario risk index, and obtaining a classification score by weighting the scores according to fault type, includes:

[0085] Step S210: Construct a two-dimensional classification system. The two-dimensional classification system includes a fault severity index and a scenario risk index. The fault severity index is used to represent the severity of the fault itself, and the scenario risk index is used to represent the risk level of the current driving scenario of the vehicle. The two-dimensional classification system includes two indicators: the fault severity index and the scenario risk index. The fault severity index is used to represent the severity of the fault itself of the vehicle, and the scenario risk index is used to represent the risk level of the current driving scenario of the vehicle. By combining these two indicators, a multi-dimensional classification judgment system is constructed.

[0086] Step S211: Assign a first preset weight to the fault severity index and a second preset weight to the scenario risk index. The first preset weight is adjusted in real time according to the fault type, and the second preset weight is adjusted in real time according to the current scenario risk level. The first preset weight will be adjusted in real time according to the fault type actually identified by the vehicle, so that the weight allocation is adapted to the characteristics of the fault itself. The second preset weight will be adjusted in real time according to the risk level corresponding to the current driving scenario of the vehicle, so that the weight allocation fits the actual risk status of the scenario. The dynamic adjustment of the weights makes the subsequent classification calculation more targeted.

[0087] Step S212: Based on the fault type, determine the corresponding fault severity index value; combined with the real-time driving scenario of the vehicle, determine the scenario risk index value; based on the identified specific fault types of the vehicle, match and determine the corresponding value for the fault severity index, ensuring that the fault severity index value matches the actual severity of the fault; simultaneously, combined with the current real-time driving scenario of the vehicle, analyze the actual risk situation of the scenario, match and determine the corresponding value for the scenario risk index, ensuring that the scenario risk index value matches the actual risk level of the driving scenario.

[0088] Step S213: The fault severity index and its corresponding first preset weight, and the scenario risk index and its corresponding second preset weight are weighted and calculated to obtain a graded score. The fault severity index and its corresponding first preset weight are calculated, and the scenario risk index and its corresponding second preset weight are calculated simultaneously. The results of these two calculations are then combined to obtain a comprehensive graded score. This graded score comprehensively reflects the dual impact of fault severity and scenario risk level.

[0089] In one embodiment of this application, the step of determining the downgrade level in a two-dimensional grading system based on the grading score includes:

[0090] Step S220: Divide the graded score into at least three intervals, each interval corresponding to a downgrade level, and the downgrade level is positively correlated with the score interval; match the corresponding autonomous driving downgrade level to each divided score interval, the level of downgrade is positively correlated with the value of the score interval, and the score falling into the interval with the higher value corresponds to the higher downgrade level. Through the clear correspondence between intervals and levels, a judgment standard is established for the subsequent matching of downgrade levels.

[0091] Step S221: Obtain the graded score, substitute the graded score into the preset score interval, and match to obtain the corresponding preliminary result of the downgrade level; extract the actual graded score of the vehicle obtained by weighted calculation, substitute the actual graded score into the pre-divided score interval, and determine the autonomous driving downgrade level corresponding to the graded score according to the matching rule of score and interval, and obtain the preliminary result of the downgrade level.

[0092] For example, the grading score is divided into three intervals: low, medium, and high, corresponding to three levels: slight downgrade (40-60), moderate downgrade (60-80), and severe downgrade (80-100), respectively. If the calculated vehicle grading score is 68 points, this score falls within the moderate downgrade interval of 60-80 points, and the preliminary result of the current downgrade level is matched and determined as a moderate downgrade.

[0093] Step S222: Combining the assigned values ​​of the fault severity index and the scenario risk index, the preliminary result of the downgrade level is verified, and the verified downgrade level is output. After obtaining the preliminary downgrade level result, the assigned values ​​of the fault severity index and the scenario risk index are retrieved. Based on the actual values ​​of these two indicators, the preliminary downgrade level result is verified. The rationality and suitability of the preliminary result are verified through the actual values ​​of the index assignments. After verification, the determined downgrade level, which is suitable for the vehicle fault and driving scenario, is output. In simple terms, a preliminary result is calculated first, then the severity of the fault and the safety of the scenario are checked together. If they are not suitable, the level is upgraded; if they are suitable, it is maintained; and even downgrading can be performed to obtain the final downgrade level.

[0094] In one embodiment of this application, the dynamic constraints include deceleration constraints, steering constraints, and control quantity smoothing constraints;

[0095] The steps for setting dynamic constraints based on the degradation level include:

[0096] Step S310: Establish the correspondence between downgrade levels and constraint parameters; the correspondence is set according to the safety control requirements of the vehicle under different downgrade levels.

[0097] Step S320: Match the corresponding constraint parameters according to the downgrade level, wherein there is a positive correlation between the downgrade level and the strictness of the constraint parameters; according to the autonomous driving downgrade level determined after vehicle verification, match the corresponding constraint parameters from the preset parameter system. The level of downgrade is positively correlated with the strictness of the constraint parameters. The higher the downgrade level of the vehicle, the stricter the constraint parameters are for limiting the vehicle's driving control.

[0098] Step S330: Form dynamic constraints for controlling the vehicle based on constraint parameters; After matching and obtaining the constraint parameters corresponding to the current downgrade level, use the constraint parameters as the core basis and combine the dynamic operating characteristics of the intelligent connected vehicle to construct dynamic constraints for managing the vehicle's driving state. These dynamic constraints clearly define key control dimensions such as deceleration, steering, and smoothing of control quantities during vehicle driving, so that subsequent driving control operations of the vehicle are implemented within the scope of these dynamic constraints.

[0099] Step S340: Based on the verified downgrade level, update the corresponding dynamic constraints to ensure that the constraint parameters match the current downgrade level and vehicle driving state. After forming the basic dynamic constraints, the established basic dynamic constraints are updated and adjusted in real time based on the verified driving downgrade level. During the update process, the specific values ​​and control range of the constraint parameters are optimized and adjusted in conjunction with the vehicle's current actual driving state, so that the constraint parameters in the updated dynamic constraints can simultaneously meet the vehicle's current downgrade level requirements and actual driving state.

[0100] In one embodiment of this application, the step of setting dynamic constraints according to the degradation level further includes:

[0101] Step S301: Based on the current downgrade level and driving status of the vehicle, determine the cutoff frequency of the low-pass filter, and filter transient fluctuations in the vehicle control quantities based on the cutoff frequency. The control quantities include braking pressure, steering angle, and motor torque. Using the cutoff frequency as the execution standard, filter the vehicle control quantities to remove transient fluctuations in the control quantities. The vehicle control quantities include braking pressure, steering angle, and motor torque. The filtering process makes the change trend of the control quantities more stable.

[0102] Step S302: Based on the constraint parameters corresponding to the downgrade level, set a dynamic limit range. The dynamic limit range includes the maximum fluctuation range and the rate of change limit of the control quantity. The dynamic limit range clearly defines the maximum fluctuation range of the control quantity and also determines the rate of change limit of the control quantity. By setting a dynamic limit range for the control quantity, a clear and specific limit is formed on the boundary of the change of the control quantity, reducing the occurrence of excessive fluctuations in the control quantity.

[0103] Step S303: Substitute the low-pass filtered control quantity into the dynamic limit range for verification. If the control quantity exceeds the dynamic limit range, adjust the control quantity to fall within the dynamic limit range. Substitute the low-pass filtered vehicle control quantity into the pre-set dynamic limit range for compliance verification. Check the actual value and changes of the control quantity. If any relevant indicator of the control quantity is detected to exceed the defined standard of the dynamic limit range, adjust the control quantity to ensure it falls within the dynamic limit range, thus ensuring the control quantity changes within the set boundaries.

[0104] Safe parking areas are divided into multiple levels based on safety priority, with priority given to areas with the lowest risk to vehicles and surrounding traffic participants. Safety priority is determined through risk quantification assessment, and a risk assessment model is constructed by combining parameters such as obstacle distance, traffic density, and road segment risk to achieve the selection of safe parking areas.

[0105] The parking path is generated using an improved path planning algorithm that meets the requirements of dynamic feasibility, obstacle avoidance safety, and path smoothness, adapting to vehicle driving characteristics. The improved algorithm uses the shortest path length, lowest risk, and dynamic feasibility as multi-objective optimization functions, balancing parking efficiency and safety, and overcoming the limitations of traditional path planning algorithms' single-objective optimization. An improved path planning algorithm refers to an algorithm that optimizes and enhances traditional path planning algorithms for specific application scenarios.

[0106] Improved path planning algorithms mainly fall into two categories: graph search algorithms and sampling algorithms. Both categories revolve around searching for a path from the vehicle's starting point to its destination, adapting to different driving environment characteristics. Graph search algorithms abstract the vehicle's driving environment into a topological structure composed of nodes and edges, searching for the optimal path within this structure using specific search logic. These algorithms demonstrate clear path optimality and are suitable for static or low-dynamic driving environments; Dijkstra's algorithm is a commonly used example. Sampling algorithms randomly sample nodes within the vehicle's feasible driving space, connecting nodes that meet the driving requirements to form feasible paths. These algorithms excel in obstacle avoidance, are more adaptable to complex dynamic driving environments, and exhibit better real-time performance; commonly used algorithms include the fast random tree algorithm and the probabilistic road graph algorithm.

[0107] Numerical optimization and interpolation fitting algorithms are also fundamental algorithms for autonomous driving path planning. They are often used in conjunction with the first two types of algorithms, each undertaking different path planning functions. Numerical optimization algorithms transform the path planning problem into a specific mathematical optimization problem, obtaining the optimal path by solving the optimization equations. These algorithms can accurately embed the vehicle's dynamic characteristics as constraints into the planning process, ensuring the planned path highly adapts to the vehicle's actual driving characteristics. Commonly used methods include model predictive control, quadratic programming, and nonlinear programming. Interpolation fitting algorithms primarily handle path smoothing, performing interpolation and fitting on the initially planned path to eliminate abrupt changes such as sharp angles and curves, ensuring continuous changes in the vehicle's position, speed, and acceleration during driving. Commonly used methods include fifth-order polynomial interpolation and Bézier curves.

[0108] Safe parking and warning operations include smooth vehicle parking, locking the parking brake, and sending warnings and status information to surrounding road users and the monitoring platform. The warning information adopts a dual warning method of visual warning and vehicle-to-everything (V2X) communication warning to ensure that surrounding road users are aware of the warning in time and reduce the safety risks such as rear-end collisions.

[0109] In addition, to clearly demonstrate the entire process from fault identification to dynamic constraint setting, taking a certain intelligent connected vehicle driving autonomously on a highway as an example, firstly, multi-source data on the vehicle is collected in real time. The lidar point cloud data is intermittent, the feedback value of the steering angle sensor deviates from the command value, and the motor torque exceeds the normal threshold range for 3 consecutive seconds.

[0110] Subsequently, the continuity features (frame drop rate of 15%), consistency features (root mean square deviation of command and feedback of 2.3%), and time-series trend features (cumulative torque deviation continuously increases) of the data were extracted. After dynamic weight calibration, the feature comprehensive value was calculated and exceeded the preset threshold. Moreover, two of the three features (continuity and time-series trend) were abnormal at the same time. The joint judgment rule thus identified the real fault type as "partial failure of the perception and execution system".

[0111] Next, based on the fault type, the fault severity index was assigned a score of 75 (out of 100), and combined with the high traffic density of the current highway scenario, the scenario risk index was assigned a score of 80. After weighted calculation, the grade score was 78, falling into the "moderate downgrade" range. After verification, the downgrade level was determined to be Level 2. Finally, based on the Level 2 downgrade, corresponding dynamic constraints were set: the deceleration upper limit was adjusted from -3m / s² to -2m / s², the steering angle rate was limited to 15 degrees / second, and a low-pass filter (cutoff frequency 2Hz) was introduced to filter transient fluctuations in the control quantity. At the same time, the control quantity was verified in conjunction with the dynamic amplitude limit range (brake pressure change rate ≤ 8MPa / s) to ensure the smoothness and safety of subsequent path planning and parking control.

[0112] like Figure 2 As shown, this application also provides an intelligent connected vehicle autonomous driving control system, which includes: an identification module 10, a calculation module 20, a search module 30, and a control module 40.

[0113] The identification module 10 is used to collect multi-source vehicle data, extract the continuity features, consistency features, and temporal trend features of the data from the multi-source data, and identify the fault type by combining the preset joint judgment rules. By collecting multi-source vehicle data such as perception, execution, and communication data in real time, the module extracts the continuity features that reflect the stability of data updates, the consistency features that reflect the matching degree of instructions and feedback, and the temporal trend features that characterize the accumulation characteristics of abnormal data from these data, and fuses these three types of features. Then, the identification module 10 cross-validates the fused features according to the preset joint judgment rules. When at least two features are abnormal at the same time, it is determined to be a real fault and the specific fault type is identified. If only a single feature is abnormal, it is determined to be noise interference and is not triggered. In this way, high-precision fault identification is achieved in the early stage of fault occurrence, effectively eliminating misjudgment or missed judgment caused by external interference.

[0114] The calculation module 20 is used to construct a two-dimensional classification system of fault severity index and scenario risk index. It calculates a classification score by weighting the fault type and determines the downgrade level in the two-dimensional classification system based on the classification score. By constructing a two-dimensional classification system that includes fault severity index and scenario risk index, the fault severity index is assigned a value based on the fault type output by the identification module 10, and the scenario risk index is assigned a value based on the risk status of the current driving scenario of the vehicle. The calculation module 20 assigns dynamic weights to these two indicators, which can be adjusted in real time according to the fault type and scenario risk level. The classification score is obtained by weighted calculation, and the score is then substituted into a preset score range to obtain a preliminary result of the downgrade level. Finally, the preliminary result is verified by combining the actual values ​​of the two indicators, and a downgrade level that is highly adapted to the actual fault state and driving environment of the vehicle is output.

[0115] The search module 30 is used to set dynamic constraints according to the degradation level, search for safe parking areas in real time, and generate parking paths that meet the dynamic constraints and obstacle avoidance requirements. Based on the degradation level determined by the calculation module 20, it sets matching dynamic constraints, including deceleration constraints, steering constraints, and control quantity smoothing constraints, and introduces low-pass filtering and dynamic amplitude limiting mechanisms to ensure the stability of the control quantity. The search module 30 detects the vehicle's surrounding environment in real time, searches for safe parking areas that meet safety requirements, and uses an improved multi-objective path planning algorithm to generate a parking path that simultaneously meets dynamic constraints, obstacle avoidance requirements, and path smoothness. This path achieves the best balance between path length, driving risk, and dynamic feasibility, ensuring that the vehicle can smoothly and safely drive towards the target parking area.

[0116] The control module 40 controls the vehicle to travel along the parking path to the safe parking area and performs safe parking and warning operations. Based on the parking path generated by the search module 30, it controls the vehicle's braking, steering, and drive systems to guide the vehicle smoothly along the planned path to the designated safe parking area. After the vehicle reaches the target location, the control module 40 performs safe parking operations, including smooth parking and parking brake locking. At the same time, it activates a dual warning mechanism, sending vehicle malfunction status information to surrounding traffic participants and the monitoring platform through visual warnings (such as hazard lights) and vehicle-to-everything (V2X) communication, ensuring that the vehicle can park properly after a malfunction and effectively warn the surrounding environment, reducing the risk of secondary accidents.

[0117] Other embodiments of the intelligent connected vehicle autonomous driving control system of this application refer to the above-described intelligent connected vehicle autonomous driving control method, and will not be repeated here.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for controlling autonomous driving in intelligent connected vehicles, characterized in that, The control method includes: Collect multi-source data from the vehicle, extract the continuity features, consistency features and time-series trend features of the data from the multi-source data, and identify the fault type by combining the preset joint judgment rules; A two-dimensional classification system of fault severity index and scenario risk index is constructed. A classification score is obtained by weighted calculation according to the fault type. The downgrade level is determined in the two-dimensional classification system based on the classification score. Based on the degradation level, dynamic constraints are set, safe parking areas are searched in real time, and parking paths that meet the dynamic constraints and obstacle avoidance requirements are generated. Control the vehicle to drive along the parking path to the safe parking area, and perform safe parking and warning operations.

2. The control method according to claim 1, characterized in that, The vehicle-mounted multi-source data includes perception data, execution data, and communication data. The perception data includes observation data from various vehicle-mounted sensors. The execution data includes working status data of vehicle-mounted actuators and vehicle operating parameters. The communication data includes vehicle-to-everything (V2X) communication data and operating status data of the vehicle-mounted computing platform. The steps for collecting multi-source data from vehicles include: Determine the collection range of vehicle-mounted multi-source data, and synchronously collect vehicle-mounted multi-source data based on the collection range; The collected data are categorized by data type.

3. The control method according to claim 2, characterized in that, After the steps of collecting multi-source data from the vehicle, the following are included: The multi-source data is preprocessed, the filtering threshold is adjusted based on the data type, invalid data is removed according to the filtering threshold, and the data format is standardized.

4. The control method according to claim 1, characterized in that, The continuity feature is used to represent the stability of data updates, the consistency feature is used to represent the matching degree between multi-source data instructions and feedback, and the time-series trend feature is used to represent the cumulative characteristics of abnormal data. The steps for extracting the continuity features, consistency features, and time-series trend features of the data from the multi-source data include: The continuity feature, consistency feature, and time-series trend feature of the data are extracted separately to obtain the continuous feature extraction value, consistency feature extraction value, and time-series trend feature extraction value. Initial extraction weights are assigned to the continuity feature, the consistency feature, and the time-series trend feature, wherein the initial weights are preset according to the multi-source data type; The initial extraction weights are dynamically calibrated based on the real-time driving scenario of the vehicle. The calibrated initial extraction weights are weighted and calculated with the corresponding feature extraction values ​​to obtain the fused feature comprehensive value.

5. The control method according to claim 1, characterized in that, The steps for identifying fault types by combining preset joint determination rules include: The system determines the anomalies based on a preset threshold and duration, wherein the preset threshold is updated in real time according to the vehicle's driving status and environmental changes. When at least two features are abnormal at the same time, it is determined to be a real fault and the fault type is identified; When a single feature is abnormal, it is judged as noise interference and fault type identification is not triggered.

6. The control method according to claim 1, characterized in that, The steps for constructing a two-dimensional classification system of fault severity index and scenario risk index, and obtaining a classification score by weighting the scores according to the fault types, include: A two-dimensional classification system is constructed, which includes a fault severity index and a scenario risk index. The fault severity index is used to represent the severity of the fault itself, and the scenario risk index is used to represent the risk level of the vehicle's current driving scenario. The fault severity index is assigned a first preset weight, and the scenario risk index is assigned a second preset weight, wherein the first preset weight is adjusted in real time according to the fault type, and the second preset weight is adjusted in real time according to the current scenario risk level. Based on the fault type, determine the corresponding fault severity index value, and combine it with the real-time driving scenario of the vehicle to determine the scenario risk index value. The severity of the fault index and its corresponding first preset weight, as well as the scenario risk index and its corresponding second preset weight, are weighted and calculated to obtain a graded score.

7. The control method according to claim 6, characterized in that, The steps for determining the downgrade level in the two-dimensional grading system based on the grading score include: The graded score is divided into at least three intervals, each interval corresponding to a downgrade level, and the downgrade level is positively correlated with the score interval; The graded score is obtained, and the graded score is substituted into a preset score range to obtain the preliminary result of the corresponding downgrade level. By combining the assigned values ​​of the fault severity index and the scenario risk index, the preliminary result of the downgrade level is verified, and the verified downgrade level is output.

8. The control method according to claim 7, characterized in that, The dynamic constraints include deceleration constraints, steering constraints, and control quantity smoothing constraints; The steps for setting dynamic constraints based on the degradation level include: Establish the correspondence between the degradation levels and constraint parameters; Based on the degradation level, a corresponding constraint parameter is matched, wherein the degradation level is positively correlated with the strictness of the constraint parameter; The dynamic constraints for controlling the vehicle are formed based on the aforementioned constraint parameters; Based on the verified downgrade level, the corresponding dynamic constraints are updated to match the constraint parameters with the current downgrade level and vehicle driving state.

9. The control method according to claim 8, characterized in that, The step of setting dynamic constraints based on the degradation level further includes: Based on the vehicle's current downgrade level and driving status, the cutoff frequency of the low-pass filter is determined, and transient fluctuations in the vehicle control quantities are filtered based on the cutoff frequency. The control quantities include braking pressure, steering angle, and motor torque. Based on the constraint parameters corresponding to the downgrade level, a dynamic limiting range is set, wherein the dynamic limiting range includes the maximum fluctuation range and the rate of change limit of the control quantity; The control quantity filtered by the low-pass filter is substituted into the dynamic limiting range for verification. If the control quantity exceeds the dynamic limiting range, the control quantity is adjusted to the dynamic limiting range.

10. An intelligent connected vehicle autonomous driving control system, characterized in that, The control system includes: The identification module is used to collect multi-source data from the vehicle, extract the continuity features, consistency features and time-series trend features of the data from the multi-source data, and identify the fault type by combining the preset joint judgment rules. The calculation module is used to construct a two-dimensional classification system of fault severity index and scenario risk index, perform weighted calculation according to the fault type to obtain a classification score, and determine the downgrade level in the two-dimensional classification system based on the classification score; The search module is used to set dynamic constraints according to the degradation level, search for safe parking areas in real time, and generate parking paths that meet the dynamic constraints and obstacle avoidance requirements. The control module is used to control the vehicle to travel along the parking path to the safe parking area and to perform safe parking and warning operations.

Images