Dynamic fault-tolerant control method and system based on expected function security mechanism

Abstract

The invention provides a dynamic fault-tolerant control method based on an expected function safety mechanism, and the method comprises the steps: 1, selecting the model of a sensor, and enabling the sensor to be distributed at the periphery of a vehicle body; 2, building a decision algorithm model, carrying out environment scene recognition, planning a driving path and speed according to data obtained by a sensing layer, evaluating safety risks of different decisions, and carrying out judgment and decision making on various conditions; 3, control interfaces of a vehicle power system, a steering system and a braking system are established, and the electronic control unit converts a decision control instruction output by the decision algorithm model into actual vehicle operation through the control interfaces; 4, performing an expected function safety mechanism, performing real-time monitoring on state parameters of the vehicle, and analyzing detection data through a fault detection algorithm; 5, implementing a fault tolerance strategy; and 6, continuously monitoring the actual driving state and the system response condition of the vehicle, and adjusting the fault tolerance strategy.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to a dynamic fault-tolerant control method and system based on an expected functional safety mechanism. Background Art

[0002] With the advancement of science and technology, intelligent driving technology has been widely adopted. Many intelligent vehicles utilize fault-tolerant control technology to improve operational comfort. However, existing technologies typically rely on a "two-tier monitoring architecture (component-level and vehicle-level only)" with fixed resource allocation (based on Infineon's TC397) and purely on-board decision-making. Alternatively, these approaches employ "offline scenario playback testing + pure software simulation verification + risk warning only without control intervention," or "functional safety-driven redundant design + pure vision system with no sensor cross-validation + no external communication compensation mechanism." However, these technical solutions lack sufficient architectural layer coverage. Monitoring units are limited to the decision-making layer, often focusing on a single level. This leads to rigid architectures and poor dynamic adaptability. They also lack real-time dynamic processing and rely on pre-built scenario libraries, making them unable to cope with sudden traffic fluctuations. Furthermore, resource allocation is rigid and computing power utilization is low. Summary of the Invention

[0003] In view of this, the present invention provides a dynamic fault-tolerant control method and system based on the expected functional safety mechanism to solve the technical problems of the existing technology such as single system architecture, limited risk control, and low resource utilization.

[0004] The present invention provides a dynamic fault-tolerant control method based on an expected functional safety mechanism, which is applied to an autonomous driving vehicle. The method comprises: step 1, selecting a sensor model according to the vehicle size, and distributing the sensors around the vehicle body to form a perception layer; step 2, building a decision algorithm model, performing environmental scene recognition through a deep learning model, a path planning algorithm, a risk assessment algorithm, and a decision algorithm, planning a driving path and speed based on the data obtained by the perception layer, evaluating the safety risks of different decisions, and making judgments and decisions on various situations; step 3, establishing a control interface for the vehicle power system, steering system, and braking system so that the electronic control unit can control the vehicle according to the functional formula through the control interface. , converting the decision control instructions output by the decision algorithm model into actual vehicle operations, where is the lateral velocity, is the yaw angular velocity, / is the lateral force of the front / rear tires, m is the vehicle mass, is the moment of inertia; step 4, perform the expected functional safety mechanism, monitor the vehicle's status parameters in real time, and analyze the monitoring data through the fault detection algorithm; step 5, when the monitoring data is analyzed to be abnormal, the decision algorithm model implements the corresponding fault tolerance strategy according to the fault type and severity; step 6, when executing the fault tolerance strategy, continuously monitor the actual driving status of the vehicle and the system response, compare the real-time feedback data with the expected control target, and when there is a deviation between the two, adjust the fault tolerance strategy to achieve adjustment and optimization of the fault tolerance strategy.

[0005] Furthermore, the method also includes: step 7, evaluating vehicle performance indicators, then conducting simulation evaluation, and then conducting actual road test verification, and optimizing and improving according to the actual road test results.

[0006] Furthermore, the step 1 includes: step 11, selecting the model of camera, millimeter radar, and laser radar according to the size of the vehicle; step 12, setting cameras in the front, sides, and rear of the vehicle, and the formula for target description using the information collected by the cameras is: , multiple millimeter-wave radars are set up around the car body, and a laser radar is set up on the roof. The formula for target description based on the information collected by the millimeter-wave radar is: , thus forming a full range of perception coverage, among which, is the image pixel coordinate, is the position of the target in the camera coordinate system, K is the camera internal parameter matrix, r is the measured distance, is the target position, is Gaussian noise; in step 13, a data fusion algorithm is used to fuse the data collected by multiple sensors; in step 14, the data of different sensors are unified into the same coordinate system through time synchronization and space calibration, so as to comprehensively analyze the advantages of each sensor data and improve the recognition accuracy and reliability of the target object.

[0007] Furthermore, the step 2 includes: step 21, processing the environmental scene recognition through the deep learning model; step 22, combining the fused sensor data with the current state of the vehicle according to the path planning algorithm, according to the following function: , , plan a reasonable driving path and speed, where N p is the prediction time domain, Q, R are weight matrices, s is the current state, For action, is the state transition probability; step 23, using a risk assessment algorithm to evaluate the security risks of different decisions; step 24, during the operation of the decision algorithm, based on the information in the preset knowledge base and rule base, judge and make decisions on various situations.

[0008] Furthermore, the step 2 also includes: step 20, collecting and organizing traffic regulations, common driving scenario cases and safe driving experience, and building a knowledge base and rule base.

[0009] Furthermore, the vehicle status parameters include vehicle speed, acceleration, steering angle, tire pressure, and battery power.

[0010] Furthermore, the method for analyzing the detection data by the fault detection algorithm includes: step 41, a fault detection method based on a model, collecting normal driving data, fault data and edge case data, and after pre-processing the data, according to the function , , compare the actual monitoring data with the model output, when , determine that the system has failed, is the residual mean, is the residual standard deviation, k is the confidence coefficient; Step 42, based on the data-driven fault detection method, a large amount of historical data is learned using a machine learning algorithm to establish a fault pattern recognition model to identify the fault characteristics in the current data in real time; Step 43, the cause of the fault is gradually investigated through fault tree analysis, and the top event is decomposed downward using logic gates until the bottom basic event, thereby narrowing the fault scope and Calculate the probability of occurrence of the top event, and take measures to isolate the faulty component from the system for isolable faults to prevent the fault from spreading and affecting the normal operation of other components. is the probability of the top event occurring, is the minimum fault combination that leads to a given event, and i is the number of fault combinations; step 44, assessing the severity of the fault based on the fault type, fault impact and safety risk.

[0011] Furthermore, the method for implementing the corresponding fault tolerance strategy is: when the fault type is a sensor failure, switch to a backup sensor or use a data fusion algorithm to replace it with other sensor data; when the fault type is an actuator failure, adjust the control strategy and use redundant actuators to maintain the basic control function of the vehicle.

[0012] Furthermore, step 5 also includes: when environmental risks or abnormal vehicle status are monitored, the decision algorithm model formulates a risk avoidance strategy based on the risk level and the actual situation of the vehicle, and when an extremely dangerous situation is monitored or a serious failure occurs in the system and normal driving cannot be maintained, an emergency safety strategy is activated.

[0013] The present invention also provides a system for implementing the dynamic fault-tolerant control method based on the expected functional safety mechanism described in claims 1-9, characterized in that the system includes: a multi-source sensor array, distributed around the vehicle body, for collecting data to form a perception layer; a judgment and decision-making module, connected to the multi-source sensor array, for building a decision-making algorithm model through a deep learning model, a path planning algorithm, a risk assessment algorithm, and a decision-making algorithm, for performing environmental scene recognition, planning the driving path and speed according to the data obtained by the perception layer, evaluating the safety risks of different decisions, making judgments and decisions on various situations, establishing a control interface for the vehicle power system, steering system and braking system, and the electronic control unit controls the vehicle through The interface converts the decision control instructions output by the decision algorithm model into actual vehicle operations, performs the expected functional safety mechanism, monitors the vehicle's status parameters in real time, and analyzes the detection data through the fault detection algorithm; the fault tolerance module is connected to the judgment and decision module. When abnormal data is monitored, the decision algorithm model implements the corresponding fault tolerance strategy according to the fault type and severity. When executing the fault tolerance strategy, it continuously monitors the actual driving status and system response of the vehicle, compares the real-time feedback data with the expected control target, and when it is found that the actual status deviates from the expected target, the fault tolerance strategy is adjusted in time, thereby realizing the adjustment and optimization of the fault tolerance strategy.

[0014] The present invention provides a dynamic fault-tolerant control method and system based on an expected functional safety mechanism. This technical solution blocks fault propagation through multiple defense layers. When any link fails, a supplementary mechanism is automatically triggered. A central safety coordinator (CSC) is designed to integrate the analysis results of two safety standards. A dynamic scenario generation engine based on reinforcement learning (real-time injection of V2X traffic flow data) and a GRU neural network are used to predict the intentions of traffic participants. A hardware-accelerated spatiotemporal alignment module and a hierarchical response strategy are used to correspond to different control degradation speeds. Dynamic computing power allocation is also implemented. A confidence-based sensor wake-up strategy and hardware-level cross-validation are used to reduce the power consumption of millimeter-wave radar and camera pixel alignment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of a dynamic fault-tolerant control method based on an expected functional safety mechanism provided by the present invention; Figure 2 This is another flow chart of a dynamic fault-tolerant control method based on an expected functional safety mechanism provided by the present invention; Figure 3 This is a flow chart of a method for constructing a perception layer provided by the present invention; Figure 4 This is a flow chart of a method for building a decision algorithm model provided by the present invention; Figure 5This is a flow chart of a detection data analysis method provided by the present invention. DETAILED DESCRIPTION

[0016] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0017] Example 1: The present invention provides a dynamic fault-tolerant control method and system based on an expected functional safety mechanism, wherein the system comprises a multi-source sensor array, a judgment and decision module, and a fault-tolerant module. Figure 1 As shown, the operating steps of the system are as follows.

[0018] Step 1: Select sensor models based on vehicle size and distribute sensors around the vehicle body to form a perception layer. A multi-source sensor array is distributed around the vehicle body to collect data and form a perception layer.

[0019] Step 2: Build a decision-making algorithm model using deep learning models, path planning algorithms, risk assessment algorithms, and decision-making algorithms. This model is used to identify environmental scenarios, plan driving paths and speeds based on data obtained from the perception layer, assess the safety risks of different decisions, and make judgments and decisions on various situations. Step 3: Establish the control interface of vehicle power system, steering system and braking system. The electronic control unit controls the vehicle through the control interface according to the function. Convert the decision control instructions output by the decision algorithm model into actual vehicle operations, where: is the lateral velocity, is the yaw angular velocity, / is the lateral force of the front / rear tires, m is the vehicle mass, is the moment of inertia; Step 4: Implement the expected functional safety mechanism, monitor the vehicle's status parameters in real time, and analyze the detection data using the fault detection algorithm; The judgment and decision-making module is connected to the multi-source sensor array and is used to build a decision-making algorithm model through deep learning models, path planning algorithms, risk assessment algorithms, and decision-making algorithms. It is used to identify environmental scenes, plan driving paths and speeds based on data obtained from the perception layer, evaluate the safety risks of different decisions, make judgments and decisions on various situations, and establish control interfaces for the vehicle's power system, steering system, and braking system. The electronic control unit converts the decision control instructions output by the decision algorithm model into actual vehicle operations through the control interface, performs expected functional safety mechanisms, monitors the vehicle's status parameters in real time, and analyzes the detection data through the fault detection algorithm.

[0020] Step 5: When data anomalies are detected, the decision-making algorithm model implements the corresponding fault tolerance strategy based on the fault type and severity; Step 6: When executing the fault tolerance strategy, continuously monitor the actual driving status of the vehicle and the system response, compare the real-time feedback data with the expected control target, and adjust the fault tolerance strategy in a timely manner when a deviation is found between the actual status and the expected target, thereby achieving adjustment and optimization of the fault tolerance strategy.

[0021] The fault tolerance module is connected to the judgment and decision-making module. When data anomalies are detected, the decision-making algorithm model implements the corresponding fault tolerance strategy based on the fault type and severity. When executing the fault tolerance strategy, it continuously monitors the actual driving status of the vehicle and the system response, and compares the real-time feedback data with the expected control target. When a deviation is found between the actual status and the expected target, the fault tolerance strategy is adjusted in time, thereby realizing the adjustment and optimization of the fault tolerance strategy.

[0022] This application provides a dynamic fault-tolerant control method and system based on an expected functional safety mechanism. By constructing a perception layer, a decision layer, and an execution layer, this system uses a multi-level defense system to block fault propagation, automatically trigger compensation mechanisms, and perform real-time safety monitoring, fault diagnosis and analysis, safety policy execution, fault pattern identification and classification, fault-tolerant control strategy generation, control strategy adjustment and optimization, and system performance evaluation and verification to improve dynamic fault-tolerant control. This technical solution improves the technical performance of intelligent driving vehicle control systems, achieves a leap in safety levels, and increases reliability in the minimum risk state.

[0023] Example 2: The present invention provides a dynamic fault-tolerant control method and system based on an expected functional safety mechanism, wherein the system comprises a multi-source sensor array, a judgment and decision module, and a fault-tolerant module. Figure 2 As shown, the operating steps of the system are as follows.

[0024] Step 1: Select sensor models based on vehicle size and distribute sensors around the vehicle body to form a perception layer. This step builds the perception layer through sensor selection and layout, as well as sensor data fusion. Sensor selection and layout means that cameras, millimeter-wave radars, lidars and other sensors are reasonably distributed around the vehicle body. Different sensors complement each other to form 360-degree all-round perception coverage, providing the system with comprehensive and accurate environmental information; sensor data fusion, for example, when the camera recognizes the outline of an object, the millimeter-wave radar confirms its distance and speed, and the lidar adds accurate three-dimensional position information. After fusion, the attributes and status of the object can be determined more accurately, such as Figure 3 As shown, the step 1 includes: Step 11: Select the camera, millimeter radar, and lidar models based on the vehicle size; Step 12: Set up cameras in front, on both sides, and at the rear of the vehicle. The formula for describing the target using the information collected by the cameras is: , multiple millimeter-wave radars are set up around the car body, and a laser radar is set up on the roof. The formula for target description based on the information collected by the millimeter-wave radar is: , thus forming a full range of perception coverage, among which, is the image pixel coordinate, is the position of the target in the camera coordinate system, K is the camera internal parameter matrix, r is the measured distance, is the target position, is Gaussian noise; Step 13: Using data fusion algorithm to fuse the data collected by multiple sensors; In step 14, the data from different sensors are unified into the same coordinate system through time synchronization and spatial calibration, so as to comprehensively analyze the advantages of the data from each sensor and improve the recognition accuracy and reliability of the target object.

[0025] This step establishes a physical layer defense loop. By performing pixel-level spatiotemporal alignment of the lidar point cloud and camera image (the lidar and camera are accelerated by the FPGA to improve response speed), and real-time monitoring of the millimeter-wave radar transmission power, it eliminates the chain reaction caused by single-point failures and improves system reliability.

[0026] Step 2: Build a decision-making algorithm model using deep learning models, path planning algorithms, risk assessment algorithms, and decision-making algorithms. This model is used to identify environmental scenarios, plan driving paths and speeds based on data obtained from the perception layer, assess the safety risks of different decisions, and make judgments and decisions on various situations. Building a decision-making algorithm model means building a decision-making layer and developing a decision-making model based on the combination of deep learning and traditional algorithms. Figure 4As shown, the step 2 includes: Step 20: Collect and organize traffic regulations, common driving scenario cases, and safe driving experience to build a knowledge base and rule base.

[0027] To successfully execute the following steps, a knowledge base and rule base must be built. This data is collected and organized into information such as traffic regulations, common driving scenarios, and safe driving experiences. During operation, the decision-making algorithm uses this information to make judgments and decisions in various situations. For example, at an intersection, the vehicle determines whether to proceed, slow down, or stop and wait based on the traffic light status and intersection traffic rules, combined with information about other vehicles and pedestrians acquired by the perception layer.

[0028] Step 21, processing environmental scene recognition through deep learning model; Deep learning models are used to process complex environmental scene recognition. For example, neural networks are trained through large amounts of labeled data to enable them to accurately identify various elements and potential risks in different traffic scenarios.

[0029] Step 22: Based on the path planning algorithm, the fused sensor data is combined with the current state of the vehicle and the following function is used: , , Plan a reasonable driving path and speed, where N p is the prediction time domain, Q, R are weight matrices, s is the current state, For action, is the state transition probability; Step 23, using a risk assessment algorithm to assess the security risks of different decisions; Traditional algorithms, such as path planning algorithms and risk assessment algorithms, plan reasonable driving paths and speeds based on the data fused by the perception layer and the current state of the vehicle, and evaluate the safety risks of different decisions. Step 24: During the operation of the decision-making algorithm, judgments and decisions are made on various situations based on the information in the preset knowledge base and rule base.

[0030] This step increases the recognition rate of unknown risk scenarios from the traditional 72% to 89%, while reducing computing power consumption. The computing power consumption of the traditional fixed allocation method is 45W, while the computing power consumption of the technical solution adopted in this application is 28W.

[0031] Step 3: Establish the control interface of vehicle power system, steering system and braking system. The electronic control unit controls the vehicle through the control interface according to the function. , converting the decision control instructions output by the decision algorithm model into actual vehicle operations, where is the lateral velocity, is the yaw angular velocity, / is the lateral force of the front / rear tires, m is the vehicle mass, is the moment of inertia; Control interfaces with the vehicle's powertrain, steering system, braking system, and other systems are established and deployed through the execution layer. The electronic control unit (ECU) converts control commands output by the decision layer into actual vehicle operations. For example, the powertrain ECU receives acceleration or deceleration commands to adjust engine output power or control motor torque; the steering system ECU controls the steering angle based on steering commands; and the braking system ECU performs braking operations, achieving vehicle acceleration, deceleration, steering, and braking control. Actuator redundancy is designed to improve the reliability and fault tolerance of the execution layer, and redundancy is employed for key actuators. For example, a dual braking circuit design ensures that if one brake circuit fails, the other circuit can still ensure a certain degree of braking capability. The steering system can be equipped with a backup steering motor or mechanical steering backup, which can switch to backup mode to maintain vehicle steering function in the event of electronic steering system failure.

[0032] Step 4: Implement the expected functional safety mechanism, monitor the vehicle's status parameters in real time, and analyze the detection data using the fault detection algorithm; Step 5: When data anomalies are detected, the decision-making algorithm model implements the corresponding fault tolerance strategy based on the fault type and severity; Step 6: When executing the fault tolerance strategy, continuously monitor the actual driving status of the vehicle and the system response, compare the real-time feedback data with the expected control target, and adjust the fault tolerance strategy in a timely manner when a deviation is found between the actual status and the expected target, thereby achieving adjustment and optimization of the fault tolerance strategy.

[0033] Step 7: Evaluate vehicle performance indicators, conduct simulation evaluation, and then conduct actual road test verification. Optimize and improve based on the actual road test results.

[0034] System performance evaluation and verification is based on performance metrics. During the system design phase, specific indicators for evaluating the dynamic fault-tolerant control method are identified, such as vehicle safety (crash risk, loss of control probability), handling stability (yaw rate, roll angle), ride comfort (acceleration rate, vibration amplitude), and system fault tolerance (fault recovery time, fault tolerance). Simulation evaluation is then conducted, with extensive simulation testing of the dynamic fault-tolerant control method performed in a virtual simulation environment. By simulating various fault scenarios and driving conditions, performance data is collected under different conditions to comprehensively evaluate the control method's performance. Simulation results are analyzed to identify shortcomings in the control method and provide a basis for further optimization. Validation through actual road testing is still required. After completing the simulation evaluation, actual road testing is conducted. Various monitoring devices are installed on the test vehicle to record the execution of the dynamic fault-tolerant control method and the vehicle's performance when encountering faults during actual driving. Comparison and analysis of actual test data with simulation results further validates the control method's effectiveness and reliability in real-world environments. Based on the actual road test results, the control method is finally optimized and improved to ensure that it can meet the actual application requirements of L3 autonomous driving.

[0035] This application provides a dynamic fault-tolerant control method and system based on an expected functional safety mechanism. By constructing a perception layer, a decision layer, and an execution layer, this system uses a multi-level defense system to block fault propagation, automatically trigger compensation mechanisms, and perform real-time safety monitoring, fault diagnosis and analysis, safety policy execution, fault pattern identification and classification, fault-tolerant control strategy generation, control strategy adjustment and optimization, and system performance evaluation and verification to improve dynamic fault-tolerant control. This technical solution improves the technical performance of intelligent driving vehicle control systems, achieves a leap in safety levels, and increases reliability in the minimum risk state.

[0036] Example 3: The present invention provides a dynamic fault-tolerant control method and system based on an expected functional safety mechanism, wherein the system comprises a multi-source sensor array, a judgment and decision module, and a fault-tolerant module. Figure 1 As shown, the operating steps of the system are as follows.

[0037] Step 1: Select sensor models based on vehicle size and distribute sensors around the vehicle body to form a perception layer. Step 2: Build a decision-making algorithm model using deep learning models, path planning algorithms, risk assessment algorithms, and decision-making algorithms. This model is used to identify environmental scenarios, plan driving paths and speeds based on data obtained from the perception layer, assess the safety risks of different decisions, and make judgments and decisions on various situations. Step 3: Establish the control interface of vehicle power system, steering system and braking system. The electronic control unit controls the vehicle through the control interface according to the function. , converting the decision control instructions output by the decision algorithm model into actual vehicle operations, where is the lateral velocity, is the yaw angular velocity, / is the lateral force of the front / rear tires, m is the vehicle mass, is the moment of inertia; Step 4: Implement the expected functional safety mechanism, monitor the vehicle's status parameters in real time, and analyze the detection data using the fault detection algorithm; The intended functional safety mechanism primarily involves real-time safety monitoring, fault diagnosis and analysis, and safety policy execution. Real-time safety monitoring in this step primarily focuses on vehicle status monitoring, continuously monitoring various vehicle parameters, including speed, acceleration, steering angle, tire pressure, and battery charge. Sensors collect this data in real time and transmit it to the decision-making layer for analysis. For example, the system can detect abnormal changes in speed or steering angles that differ from expectations, potentially indicating a vehicle malfunction or unstable driving conditions. Environmental risk monitoring also involves monitoring potential risks in the vehicle's surroundings based on sensor-layer data, identifying dangerous situations such as sudden braking or lane changes by vehicles ahead, illegal pedestrian crossings, and obstacles on the road. Risk assessment algorithms are used to quantify these potential risks and determine their level. For example, detecting a close approach to a vehicle ahead with sudden braking signals indicates a high-risk state, triggering appropriate safety countermeasures. Finally, system component monitoring monitors the operating status of various system components, such as sensors, controllers, and actuators. Self-test programs and fault diagnosis algorithms are used to determine in real time whether components are functioning properly. For example, regularly check whether the sensor data output is within a reasonable range, whether the controller calculation is normal, and whether the actuator action responds in time. Once a component failure is found, the fault location and alarm are carried out in time. Fault diagnosis and analysis will use a variety of fault detection algorithms to analyze and process the monitoring data, such as Figure 5 As shown, the method for analyzing detection data using a fault detection algorithm includes: Step 41, based on the model fault detection method, collects normal driving data, fault data and edge case data, and after preprocessing the data, uses the function , , compare the actual monitoring data with the model output, when , determine that the system has failed, is the residual mean, is the residual standard deviation, k is the confidence coefficient; Step 42: Based on a data-driven fault detection method, a machine learning algorithm is used to learn a large amount of historical data, establish a fault pattern recognition model, and identify fault characteristics in the current data in real time; Step 43, through the fault tree analysis, gradually find out the cause of the fault, use the logic gate to decompose the top event down to the bottom basic event, narrow the scope of the fault, and according to the function Calculate the probability of occurrence of the top event, and take measures to isolate the faulty component from the system for isolable faults to prevent the fault from spreading and affecting the normal operation of other components. is the probability of the top event occurring, is the minimum fault combination that leads to a given event, i is the number of fault combinations; Fault location and isolation is the process of rapidly locating a fault once it is detected, identifying the specific component or subsystem where the fault occurred. For isolable faults, measures are taken to isolate the faulty component from the system to prevent the fault from spreading and impacting the normal operation of other components. For example, if a sensor fails, the system automatically switches to a backup sensor and blocks the faulty sensor's data output.

[0038] Step 44 : Evaluate the severity of the fault based on the fault type, fault impact, and safety risk.

[0039] Fault severity is assessed based on the fault type, its impact on system functionality, and the potential safety risks. Faults are categorized into different levels, such as minor, general, and severe. For severe faults, emergency response measures are immediately initiated to ensure vehicle and passenger safety. For general and minor faults, fault information is recorded and repairs are performed at the appropriate time. Furthermore, fault pattern identification and classification requires the establishment of a fault pattern library. During the system development phase, a detailed fault pattern library is developed through analysis and simulation of various possible faults. This library covers various fault modes, including sensor faults, controller faults, actuator faults, and communication faults. Each fault mode's characteristics, causes, and potential impacts are described and recorded. Real-time fault pattern matching is performed. During system operation, when the fault diagnosis module detects a fault, it matches the fault characteristics with the information in the fault pattern library in real time to quickly identify the specific fault mode to which the fault belongs. For example, by analyzing abnormal changes in sensor data, it can determine whether the fault mode is sensor hardware damage, signal interference, or data transmission errors.

[0040] The expected functional safety in this step works in conjunction with the original functional safety to form a dual-standard assessment matrix for the Central Safety Coordinator (CSC), which can effectively reduce safety conflict incidents.

[0041] Step 5: When data anomalies are detected, the decision-making algorithm model implements the corresponding fault tolerance strategy based on the fault type and severity; The method for implementing the corresponding fault-tolerant strategy involves: when the fault type is a sensor failure, switching to a backup sensor or using a data fusion algorithm to replace it with other sensor data. When the fault type is an actuator failure, the control strategy is adjusted to utilize redundant actuators to maintain basic vehicle control functions. For example, if a steering motor fails, the system switches to the backup steering motor and adjusts the steering control algorithm to ensure the vehicle can continue to steer safely. Fault-tolerant strategy generation involves formulating a corresponding fault-tolerant control strategy based on the fault pattern recognition results, the vehicle's current driving state, and safety requirements, in accordance with pre-established fault-tolerant control strategy generation principles. Strategy development principles include ensuring basic safe driving functions, minimizing the impact of faults on vehicle performance, and prioritizing the use of redundant resources for fault tolerance. Strategy generation based on model prediction utilizes a model predictive control algorithm to predict the vehicle's future driving state under different fault-tolerant control strategies. Through simulation analysis, the feasibility and effectiveness of various strategies are evaluated to select the optimal fault-tolerant control strategy. For example, in the event of an actuator failure, the system predicts the vehicle's trajectory, stability, and handling performance under different redundant actuator combinations and control parameter adjustment schemes, selecting the strategy that maximizes vehicle performance while meeting safety requirements. After completing the detection and assessment process, in addition to implementing fault-tolerant strategies, the system also includes: When environmental risks or abnormal vehicle conditions are detected, the decision-making algorithm model develops risk avoidance strategies based on the risk level and the vehicle's actual conditions. For example, if an obstacle is detected ahead and cannot be bypassed in time, the system automatically reduces speed to maintain a safe distance and alerts the driver through a warning device. If the vehicle ahead suddenly brakes, the system quickly determines the distance and relative speed to the preceding vehicle and automatically applies braking measures to avoid a rear-end collision. When an extremely dangerous situation is detected or a serious system failure occurs, making it impossible to maintain normal driving, emergency safety strategies are activated, such as automatically guiding the vehicle to a safe area, such as an emergency lane or roadside safety zone. Simultaneously, a distress signal is sent to the rescue center via the communication system, notifying the vehicle's location and fault status. The vehicle doors are unlocked to facilitate passenger escape and rescue efforts.

[0042] Step 6: When executing the fault tolerance strategy, continuously monitor the actual driving status of the vehicle and the system response, compare the real-time feedback data with the expected control target, and adjust the fault tolerance strategy in a timely manner when a deviation is found between the actual status and the expected target, thereby achieving adjustment and optimization of the fault tolerance strategy.

[0043] For example, the braking force and duration of the braking control strategy are dynamically adjusted based on the vehicle's actual deceleration and the real-time distance to obstacles. Furthermore, adaptive vehicle optimization takes into account the complexity of the vehicle's driving environment and operating conditions. The fault-tolerant control strategy features adaptive optimization capabilities. Control parameters and strategies are automatically optimized based on varying road conditions (such as dry, slippery, and icy roads), traffic conditions (congested or smooth), and vehicle load. For example, on slippery roads, the vehicle's acceleration and deceleration limits are appropriately reduced to improve steering stability and adapt to reduced road adhesion. This application provides a dynamic fault-tolerant control method and system based on an expected functional safety mechanism. By constructing a perception layer, a decision layer, and an execution layer, this system uses a multi-level defense system to block fault propagation, automatically trigger compensation mechanisms, and perform real-time safety monitoring, fault diagnosis and analysis, safety policy execution, fault pattern identification and classification, fault-tolerant control strategy generation, control strategy adjustment and optimization, and system performance evaluation and verification to improve dynamic fault-tolerant control. This technical solution improves the technical performance of intelligent driving vehicle control systems, achieves a leap in safety levels, and increases reliability in the minimum risk state.

[0044] In summary, the embodiments of the present invention provide a dynamic fault-tolerant control method and system based on the expected functional safety mechanism. This technical solution improves the real-time risk response, the completeness of scenario coverage, and the efficiency of resource utilization. At the same time, it realizes the transition of safety levels, completes the coordination of functional safety and SOTIF, and improves the reliability of the minimum risk state (MRM). It also optimizes hardware costs and improves development efficiency.

[0045] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A dynamic fault-tolerant control method based on an expected functional safety mechanism, applied to an autonomous driving vehicle, characterized in that: The method comprises: Step 1: Select sensor models based on vehicle size and distribute sensors around the vehicle body to form a perception layer. Step 2: Build a decision-making algorithm model. Using deep learning models, path planning algorithms, risk assessment algorithms, and decision-making algorithms, the model identifies environmental scenarios, plans driving routes and speeds based on data acquired by the perception layer, assesses the safety risks of different decisions, and makes judgments and decisions on various situations. Step 3: Establish the control interface of vehicle power system, steering system and braking system so that the electronic control unit can control the vehicle power system, steering system and braking system according to the function through the control interface. , converting the decision control instructions output by the decision algorithm model into actual vehicle operations, where is the lateral velocity, is the yaw angular velocity, / is the lateral force of the front / rear tires, m is the vehicle mass, is the moment of inertia; Step 4: Implement the expected functional safety mechanism, monitor the vehicle's status parameters in real time, and analyze the monitoring data using the fault detection algorithm; Step 5: When abnormal monitoring data is detected, the decision-making algorithm model implements the corresponding fault tolerance strategy based on the fault type and severity; Step 6: When executing the fault tolerance strategy, continuously monitor the actual driving status of the vehicle and the system response, compare the real-time feedback data with the expected control target, and adjust the fault tolerance strategy when there is a deviation between the two, thereby achieving adjustment and optimization of the fault tolerance strategy.

2. The dynamic fault-tolerant control method based on the expected functional safety mechanism according to claim 1, characterized in that: The method further includes: step 7, evaluating vehicle performance indicators, then conducting simulation evaluation, and then conducting actual road test verification, and optimizing and improving according to the actual road test results.

3. The dynamic fault-tolerant control method based on the expected functional safety mechanism according to claim 1, characterized in that: The step 1 comprises: Step 11: Select the camera, millimeter radar, and lidar models based on the vehicle size; Step 12: Set up cameras in front, on both sides, and at the rear of the vehicle. The formula for describing the target using the information collected by the cameras is: , multiple millimeter-wave radars are set up around the car body, and a laser radar is set up on the roof. The formula for target description based on the information collected by the millimeter-wave radar is: , thus forming a full range of perception coverage, among which, is the image pixel coordinate, is the position of the target in the camera coordinate system, K is the camera internal parameter matrix, r is the measured distance, is the target position, is Gaussian noise; Step 13: Using data fusion algorithm to fuse the data collected by multiple sensors; In step 14, the data from different sensors are unified into the same coordinate system through time synchronization and spatial calibration, so as to comprehensively analyze the advantages of the data from each sensor and improve the recognition accuracy and reliability of the target object.

4. A dynamic fault-tolerant control method based on an expected functional safety mechanism according to claim 3, characterized in that: The step 2 includes: Step 21, processing environmental scene recognition through deep learning model; Step 22: Based on the path planning algorithm, the fused sensor data is combined with the current state of the vehicle and the following function is used: ， ， Plan a reasonable driving path and speed, where N p is the prediction time domain, Q, R are weight matrices, s is the current state, For action, is the state transition probability; Step 23, using a risk assessment algorithm to assess the security risks of different decisions; Step 24: During the operation of the decision-making algorithm, judgments and decisions are made on various situations based on the information in the preset knowledge base and rule base.

5. A dynamic fault-tolerant control method based on an expected functional safety mechanism according to claim 4, characterized in that: The step 2 further comprises: Step 20: Collect and organize traffic regulations, common driving scenario cases, and safe driving experience to build a knowledge base and rule base.

6. A dynamic fault-tolerant control method based on an expected functional safety mechanism according to claim 1, characterized in that: The vehicle status parameters include vehicle speed, acceleration, steering angle, tire pressure, and battery power.

7. A dynamic fault-tolerant control method based on an expected functional safety mechanism according to claim 1, characterized in that: The method for analyzing detection data by using a fault detection algorithm includes: Step 41, based on the model fault detection method, collects normal driving data, fault data and edge case data, and after preprocessing the data, uses the function , , compare the actual monitoring data with the model output, when , determine that the system has failed, is the residual mean, is the residual standard deviation, k is the confidence coefficient; Step 42: Based on a data-driven fault detection method, a machine learning algorithm is used to learn a large amount of historical data, establish a fault pattern recognition model, and identify fault characteristics in the current data in real time; Step 43, through the fault tree analysis, gradually find out the cause of the fault, use the logic gate to decompose the top event down to the bottom basic event, narrow the scope of the fault, and according to the function Calculate the probability of occurrence of the top event, and take measures to isolate the faulty component from the system for isolable faults to prevent the fault from spreading and affecting the normal operation of other components. is the probability of the top event occurring, is the minimum fault combination that leads to a given event, i is the number of fault combinations; Step 44 : Evaluate the severity of the fault based on the fault type, fault impact, and safety risk.

8. The dynamic fault-tolerant control method based on the expected functional safety mechanism according to claim 1, characterized in that: The method for implementing the corresponding fault tolerance strategy is: when the fault type is a sensor failure, switch to a backup sensor or use a data fusion algorithm to replace it with other sensor data; when the fault type is an actuator failure, adjust the control strategy and use redundant actuators to maintain the basic control function of the vehicle.

9. The dynamic fault-tolerant control method based on the expected functional safety mechanism according to claim 1, characterized in that: The step 5 also includes: when environmental risks or abnormal vehicle status are monitored, the decision algorithm model formulates a risk avoidance strategy based on the risk level and the actual situation of the vehicle, and when an extremely dangerous situation is monitored or the system has a serious failure and cannot maintain normal driving, an emergency safety strategy is activated.

10. A system for implementing the dynamic fault-tolerant control method based on the expected functional safety mechanism according to claims 1-9, characterized in that: The system comprises: A multi-source sensor array, distributed around the vehicle body, is used to collect data and form a perception layer; The judgment and decision-making module is connected to the multi-source sensor array and is used to build a decision-making algorithm model through deep learning models, path planning algorithms, risk assessment algorithms, and decision-making algorithms. It is used to identify environmental scenarios, plan driving paths and speeds based on data obtained from the perception layer, evaluate the safety risks of different decisions, make judgments and decisions on various situations, and establish control interfaces for the vehicle's power system, steering system, and braking system. The electronic control unit converts the decision control instructions output by the decision algorithm model into actual vehicle operations through the control interface, implements the expected functional safety mechanism, monitors the vehicle's status parameters in real time, and analyzes the detection data through the fault detection algorithm; The fault tolerance module is connected to the judgment and decision-making module. When data anomalies are detected, the decision-making algorithm model implements the corresponding fault tolerance strategy based on the fault type and severity. When executing the fault tolerance strategy, it continuously monitors the actual driving status of the vehicle and the system response, and compares the real-time feedback data with the expected control target. When a deviation is found between the actual status and the expected target, the fault tolerance strategy is adjusted in time, thereby realizing the adjustment and optimization of the fault tolerance strategy.

Images