Vehicle fault-tolerant control method under abnormal condition of multivariable measurement sensor

Abstract

The invention discloses a vehicle fault-tolerant control method under the abnormal condition of a multivariable measurement sensor, and belongs to the crossing field of electric digital data processing and intelligent driving vehicle control technologies. The method comprises the following steps: firstly, based on spatial-temporal correlation and physical consistency constraints of multi-source sensor data, carrying out redundancy modeling and credibility evaluation on a key state of a vehicle; then, when it is detected that the sensor is abnormal, real-time compensation and correction of failure observation are achieved through a state reconstruction model driven by redundant data; and further combining a rapid fault identification and virtual sensor switching mechanism to construct a fault-tolerant control strategy for the minimum function demand of the vehicle so as to ensure that the vehicle can still maintain basic driving and safety control capability under the condition that the functions of part of sensors are damaged. According to the invention, the real-time performance of the system is ensured, the safety and robustness of the intelligent driving vehicle in a complex environment and an abnormal condition are obviously improved, and the method is suitable for a multi-sensor fusion intelligent vehicle control system.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of electronic digital data processing and intelligent driving vehicle control technology. Specifically, it relates to a fault-tolerant control method for vehicles under abnormal conditions of multivariable measurement sensors. It involves technologies such as multi-source sensor data synchronization, unified state space modeling, observation model construction, physical consistency constraint design, spatiotemporal redundancy analysis, state reconstruction, virtual sensor generation, and fault-tolerant control strategy optimization. It is applicable to the safety control of intelligent driving vehicles with multi-sensor fusion in scenarios such as sensor failure, abnormal drift, or physical attack. Background Technology

[0002] With the rapid development of intelligent driving technology, vehicle perception systems typically integrate various types of sensors, such as LiDAR, cameras, millimeter-wave radar, and inertial measurement units (IMUs). These sensors can perceive the surrounding environment and the vehicle's own state from different modalities and dimensions, providing high-precision measurement data for vehicle decision-making and control. Their reliability directly determines driving safety.

[0003] However, in complex road environments, sensors are susceptible to external interference, physical damage, or malicious attacks, resulting in problems such as failure and abnormal drift. Existing vehicle control systems have significant shortcomings: (1) They do not fully consider sensor abnormal scenarios and lack a systematic abnormal identification and response mechanism, making it easy for a single sensor failure to cause system loss of control; (2) They do not fully exploit the spatiotemporal redundancy characteristics of multi-source sensor data, making it difficult to accurately reconstruct the vehicle state when some sensors are abnormal; (3) The control strategy is not optimized for abnormal operating conditions, making it difficult to maintain basic driving safety and controllability of the vehicle when sensor functions are damaged; (4) The mode switching lacks a smooth transition mechanism, which can easily lead to sudden changes in control commands and affect the stability of vehicle operation.

[0004] Therefore, how to ensure the safe operation of vehicles under sensor malfunctions by utilizing redundant data, reconstructing states, and adjusting control strategies has become a critical issue that intelligent driving vehicle control systems urgently need to address. This invention proposes a real-time fault-tolerant control method that integrates spatiotemporal redundant sensor data with vehicle dynamics characteristics, thereby improving the safety and robustness of intelligent driving vehicles under abnormal conditions. Summary of the Invention

[0005] The purpose of this invention is to overcome the problem that existing intelligent driving vehicle control systems cannot maintain safe operation when sensors are abnormal, and to propose a fault-tolerant control method for vehicles under abnormal conditions of multivariable measurement sensors.

[0006] The technical solution of this invention is as follows: To achieve the above objectives, the present invention provides a vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors, characterized by comprising the following steps: Step 1) Unified modeling and physical consistency constraint construction for multivariable measurement sensors Multivariate measurement data from multiple vehicle sensors are acquired, and time-aligned data from different sensors are processed to construct a synchronized multivariate measurement set. Simultaneously, key vehicle physical state vectors, independent of specific sensor types, are defined to establish a unified state-space model of the vehicle's state.

[0007] Step 2) Multivariate measurement consistency analysis and anomaly detection based on spatiotemporal redundancy

[0008] Based on the unified state-space model, observation models between multi-source sensor measurements and vehicle state are established, and sensor anomaly deviation terms are introduced. Simultaneously, a vehicle dynamics state transition model is established, and a set of physical consistency constraints that the vehicle state must satisfy is constructed. Within a preset time window, based on the temporal and spatial redundancy characteristics of multi-source sensor measurements, spatiotemporal consistency analysis is performed on the measurement results of each sensor, calculating the spatiotemporal consistency metric corresponding to each sensor. Combined with the physical consistency constraints, the reliability index of each sensor is obtained.

[0009] Step 3) Credibility-weighted state reconstruction and virtual sensor generation

[0010] Based on the reliability index of each sensor, the multi-source sensor measurements are weighted according to their reliability to construct a reconstruction estimation result of the vehicle's critical state. When the reliability of any sensor is lower than a preset threshold, a corresponding virtual sensor measurement is generated based on the reconstructed state, and the virtual measurement is used to replace the measured data of the abnormal sensor to form an effective observation set.

[0011] Step 4) Construction of Real-Time Fault-Tolerant Control Strategy Oriented to Vehicle Minimum Functional Constraints

[0012] Based on the reconstructed state and the effective observation set, a set of state constraints that meet the minimum functional operation requirements of the vehicle is constructed, and the fault-tolerant control command of the vehicle is solved under the constraints, so that the vehicle can maintain basic driving and safety control functions in the event of sensor abnormality or failure.

[0013] Step 5) Fault-tolerant mode switching and normal control recovery mechanism

[0014] Based on changes in sensor reliability and vehicle operating status, the system switches between normal control mode and fault-tolerant control mode. In fault-tolerant control mode, control parameters are reconfigured, and abnormal sensors are continuously monitored. When sensor reliability recovers to the preset recovery threshold, the system smoothly switches back to normal control mode.

[0015] Step 6) System Real-time Operation and Abnormal Closed-Loop Update Mechanism

[0016] The processing flow consisting of the above steps is executed cyclically at a preset sampling period, and the stability of the vehicle state and the persistence of sensor anomalies are determined online. Multi-source sensor data is collected in real time and the vehicle state estimate is updated. Based on sensor reliability and mode state information, the anomaly determination threshold and weight parameters are updated in a closed loop to dynamically attenuate the impact of abnormal sensor data channels. When the preset termination condition is met, relevant operating information is output.

[0017] Step 7) Overall Execution Process Management and Termination Condition Determination

[0018] The overall execution flow of the method is cyclically scheduled, and key indicators such as changes in vehicle state estimation and sensor reliability are monitored online to evaluate the effectiveness of the method. Sensor anomalies and system safety status are continuously assessed. When the vehicle state continuously violates physical constraints or reaches the execution termination time, the current method execution cycle ends, and relevant information is output for use by the upper-level decision-making system.

[0019] Compared with the prior art, the present invention has the following beneficial effects: (1) Accurate anomaly identification: Combining spatiotemporal redundancy characteristics and physical consistency constraints, a multi-dimensional consistency metric and anomaly judgment function are constructed to achieve rapid and accurate identification of sensor anomalies, providing a reliable basis for fault-tolerant control.

[0020] (2) Reliable state reconstruction: Based on the credibility weighted fusion of multi-source data and combined with the virtual sensor generation mechanism, the vehicle's key state can still be continuously and accurately estimated when some sensors are abnormal, ensuring the continuity of state information.

[0021] (3) Fault-tolerant and efficient control: The control strategy is designed with the minimum functional requirements of the vehicle in mind, prioritizing the protection of core safety indicators such as speed and yaw angle, and balancing safety and basic driving requirements when sensors malfunction.

[0022] (4) Smooth mode switching: Design a dynamic switching mechanism between normal control and fault-tolerant control modes, and avoid sudden changes by gradually integrating control commands to ensure vehicle operation stability.

[0023] (5) Strong adaptability: Through parameter closed-loop update and abnormal influence attenuation mechanism, the robustness of the system under different environments and working conditions is enhanced, and the system adapts to the dynamic changes of sensor abnormalities.

[0024] In summary, the vehicle fault-tolerant control method for abnormal conditions of multivariable measurement sensors provided by this invention can ensure that intelligent driving vehicles maintain basic driving and safety control capabilities in scenarios such as sensor failure, abnormal drift, or physical attack, providing important support for the high-reliability operation of intelligent driving systems. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0026] This invention proposes a multivariable measurement sensor anomaly detection and fault-tolerant control method for intelligent driving, used to achieve highly reliable, low-redundancy state estimation and safety control in multi-sensor environments. This method significantly improves the robustness and safety of intelligent driving systems under abnormal sensor conditions by constructing a unified state model, spatiotemporal redundancy consistency analysis, credibility-weighted state reconstruction, virtual sensor generation, and a minimum function fault-tolerant control strategy. The specific implementation steps are as follows: Step 1) Unified modeling and physical consistency constraint construction for multivariable measurement sensors (1.1) Acquisition and time alignment of multi-source sensor measurement data During vehicle operation, data is collected from multiple onboard sensors at discrete moments. The original measurement data form a multivariate measurement set.

[0027] in: ： Indicates the number of sensors; ： Indicates the first The observation vector obtained by each sensor at its own sampling time.

[0028] Because different sensors have different sampling frequencies and timestamps, the measurement data from each sensor are aligned to a unified time axis and mapped to a unified system sampling time. ,Right now

[0029]

[0030] in: : To standardize the sampling period for the system; : for the first The mapping relationship between the original timestamps of each sensor and the unified time grid.

[0031] After time alignment is completed, synchronous measurements are obtained for subsequent modeling.

[0032] (1.2) Unified multivariable definition of key physical states of vehicles

[0033] Based on vehicle kinematics and control requirements, the vehicle is defined at time... The key physical state vector is

[0034] in: : Represents the vehicle's position coordinates in three-dimensional space; : Represents the vehicle speed component; : Represents the vehicle acceleration component.

[0035] This state vector serves as a unified expression of the vehicle's operating state and is independent of the specific type and number of sensors.

[0036] (1.3) Construction of a unified observation model for multivariate measurements

[0037] Regarding the first A unified observation model is established between the measurement data of each sensor and the vehicle status:

[0038] in: ; : for the first Measurement mapping function for each sensor; : This is the measurement noise vector; : is the corresponding measurement noise covariance matrix.

[0039] Anomalies are introduced to describe the abnormal characteristics of the sensor.

[0040] in: : indicates the first Each sensor at time Abnormal deviation items.

[0041] (1.4) Establishment of vehicle dynamics state transition model

[0042] Based on the vehicle's kinematic characteristics, a discrete-time transition model of the vehicle's state is constructed.

[0043] in: ; : Represents the state transition function; : For control input; : This refers to process noise; : Process noise covariance matrix.

[0044] (1.5) Construction of vehicle physical consistency constraints

[0045] Based on vehicle physical accessibility, construct a set of physical consistency constraints.

[0046] Specifically, this includes velocity and acceleration constraints: , .

[0047] Where: v(k): represents the magnitude of the vehicle's resultant velocity; : Represents the maximum permissible speed threshold for the vehicle; a(k): Represents the magnitude of the vehicle's resultant acceleration; a max : Indicates the maximum permissible acceleration threshold for the vehicle.

[0048]

[0049]

[0050] (1.6) Output of step 1

[0051] After completing the above processing, a unified modeling result is output, including the vehicle state space model, synchronous measurement data, observation model parameters, state transition model, and physical consistency constraint set.

[0052] Step 2) Multivariate measurement consistency analysis and anomaly detection based on spatiotemporal redundancy

[0053] (2.1) Modeling of temporal redundancy consistency of multi-source sensors

[0054] Based on the unified state model, observation model, and physical consistency constraints constructed in step 1, for the first... Each sensor constructs its measurement sequence within a continuous time window.

[0055] Where: Zit(k): represents the th... Each sensor at time Time window measurement sequence; : Indicates the historical backtracking step size within the time window; : Indicates the length of the time window.

[0056] Based on predicted state Calculate the corresponding predictive observations

[0057] in: : indicates the first One sensor in Predicted observations at time; : Indicates that the vehicle is in The predicted state vector at time t.

[0058] Further construct the time residual sequence

[0059] in: : indicates the first One sensor in The time residual at any given moment (the difference between the measured value and the predicted value).

[0060] And calculate time consistency metrics

[0061] in: : indicates the first A measure of the time consistency of a sensor at time k; : Indicates averaging the residuals within the time window; : indicates the first The inverse matrix of the noise covariance matrix of each sensor (used to normalize the residuals).

[0062] (2.2) Spatial redundancy consistency modeling of multi-source sensors

[0063] For a set of sensors capable of observing the same vehicle state variable A cross-sensor spatial consistency analysis model is constructed. For any two sensors... The observed difference is defined as:

[0064] in: : indicates the first A subset of sensors with observable state variables; : indicates the first The sensor and the first Each sensor at time The observed difference vector.

[0065] Constructing a spatial consistency index based on observation noise characteristics:

[0066] in: : Indicates sensor With sensors At any moment Spatial consistency index; : Represents the transpose of the observation difference vector; : Represents the inverse matrix of the sum of the noise covariance matrices of the two sensors.

[0067] Further on the first A comprehensive evaluation of the spatial consistency of each sensor is conducted.

[0068] in: : indicates the first One sensor in A comprehensive spatial consistency index at any given time; : indicates a subset of sensors The number of sensors in the system; : indicates to Except for the first Summation is performed on all other sensors besides the one sensor.

[0069] (2.3) Construction of a joint metric for spatiotemporal redundancy consistency

[0070] By combining time consistency metrics and spatial consistency metrics, the first... Spatiotemporal consistency measure of individual sensors:

[0071] in: : indicates the first Each sensor at time A comprehensive measure of spatiotemporal consistency; : Represents the weighting coefficient for consistency in the time dimension; : Represents the weighting coefficients for spatial dimensional consistency; : Indicates the range of values ​​for the weighting coefficient.

[0072] (2.4) Auxiliary determination based on physical consistency constraints

[0073] Introducing the physical consistency constraints from step 1, for the... The state components are estimated by a single sensor. Perform constraint testing:

[0074] When constraints are violated, a physical default indicator is constructed:

[0075] in: : indicates that only the first The vehicle state vector estimated by each sensor; : indicates the first The physical consistency deviation index of each sensor; : Represents the function that takes the maximum value; it is 0 if the constraint is satisfied, and a violation value if the constraint is violated. : Represents the norm operation of a vector (usually the 2-norm).

[0076] (2.5) Sensor anomaly detection and reliability calculation

[0077] By integrating spatiotemporal consistency measures and physical consistency indices, the first... Anomaly detection function for each sensor:

[0078] in: : indicates the first One sensor in The value of the anomaly detection function at any given time (the larger the value, the more abnormal it is); : Represents the physical consistency penalty coefficient.

[0079] Calculate sensor reliability based on anomaly detection function:

[0080] in: : indicates the first Each sensor at time Credibility; : Represents an exponential function.

[0081] And set the exception detection conditions:

[0082] in: : Indicates the preset sensor confidence threshold.

[0083] When the above conditions are met, determine the first... Each sensor at time It is in an abnormal state.

[0084] (2.6) Output of step 2

[0085] The final output includes the anomaly detection result and corresponding reliability index for each sensor.

[0086] in: : Indicates time The set of credibility for all sensors.

[0087] Step 3) Credibility-weighted state reconstruction and virtual sensor generation

[0088] (3.1) Reliability-based multi-source measurement weight normalization

[0089] Based on the sensor reliability index output in step 2, the weights of the multi-source sensors involved in state reconstruction are normalized to construct a reliability weighting coefficient.

[0090] in: : indicates the first Normalized confidence weights for each sensor; The weight satisfies the normalization condition:

[0091] (3.2) Vehicle state reconstruction estimation based on unified observation model

[0092] Based on the unified state-space model and observation model constructed in step 1, the key states of the vehicle are reconstructed and estimated using confidence-weighted multi-source measurements.

[0093] in: : Indicates the time of the vehicle The reconstructed state estimation vector; : indicates the first The observation mapping matrix (or filter gain) corresponding to each sensor; Its calculation form is in: : The linearized matrix of the observation function at the current estimation point; For the first Measurement noise covariance matrix of each sensor.

[0094] (3.3) Generation of virtual sensor output corresponding to abnormal sensor

[0095] When a sensor is determined to be abnormal in step 2, that is, its reliability meets the requirement...

[0096] in: : Preset credibility threshold; Based on the reconstructed vehicle state; Then based on the reconstructed vehicle state The virtual measurement output of the sensor is generated through the corresponding observation function:

[0097] in: It is used to replace the measured data in subsequent control and judgment when the sensor is abnormal or fails.

[0098] (3.4) Adaptive switching and fusion of real and virtual measurements

[0099] Based on the sensor's reliability status, the real and virtual measurements are adaptively switched and fused to form the final usable observations:

[0100] in: : Represents the effective observations after fusion; : Represents the switching coefficient based on credibility; Its definition is

[0101] Where: 1: indicates the use of actual measurement values; 0: indicates the use of virtual measurement values.

[0102] This step ultimately outputs the vehicle state reconstruction estimation result. and the fused effective observation set .

[0103] Step 4) Construction of Real-Time Fault-Tolerant Control Strategy Oriented to Vehicle Minimum Functional Constraints

[0104] (4.1) Determination of the constraint set for the minimum functional operating state of the vehicle

[0105] Based on the vehicle reconfiguration state estimate and effective observation set output from step 3, a constraint system and control objectives are constructed around the vehicle's minimum functional operating requirements. Based on the basic driving and safety requirements that the vehicle must maintain even under sensor malfunctions, the minimum functional operating state constraint set is determined.

[0106] in: : Represents the set of minimum functional operating states of a vehicle; : Represents the physical constraints that a state must satisfy.

[0107] Furthermore, a typical form of minimum function constraint is given:

[0108]

[0109]

[0110] in: : Indicates the maximum safe speed in fault-tolerant mode; : Indicates the yaw rate of the vehicle; : Indicates the yaw stability threshold; : Indicates the upper limit of safe acceleration in fault-tolerant mode.

[0111] (4.2) Construction of fault-tolerant control objectives based on reconfiguration state

[0112] Vehicle reconfiguration state estimate output in step 3 Based on this, a fault-tolerant control objective function oriented towards minimum functional constraints is constructed:

[0113] in: : Represents the objective function value of fault-tolerant control; : Indicates the predicted state at the next moment; : Represents the reference state vector in the minimum functional operating mode; : Represents the weighted norm based on matrix Qc; : Represents the state error weighting matrix.

[0114] Reference status Based on the current driving conditions, priority should be given to ensuring the vehicle's longitudinal stability, lateral controllability, and attitude safety.

[0115] (4.3) Solving the fault-tolerant control law under constraints

[0116] Under the constraints of the vehicle dynamics model, the control input is optimized and solved:

[0117]

[0118]

[0119] in: : is the vehicle control input vector; : This is the set of control inputs allowed in the minimum functional mode, used to limit the range of changes in steering angle, driving force, or braking force; u (k): Represents the optimal fault-tolerant control command obtained under constraints; : This represents finding the value that minimizes J(k) for the control input u(k); : Represents the set of control inputs allowed in the minimum functional mode; : Represents state constraints.

[0120] (4.4) Real-time fault-tolerant control output and mode hold

[0121] The obtained optimal fault-tolerant control command It acts on the vehicle's actuators and monitors the vehicle's operating status in real time.

[0122] in: : Indicates that after a control command is applied, the vehicle... The predicted state at any given moment; : Indicates that the vehicle is in The current state at any given moment; u (k): represents the optimal fault-tolerant control instruction obtained in step 4.

[0123] The output of this step is a sequence of real-time fault-tolerant control instructions that meet the minimum functional constraints. This is a direct result of control measures to maintain safe operation of the vehicle in the event of sensor malfunction or failure.

[0124] Step 5) Fault-tolerant mode switching and normal control recovery mechanism

[0125] (5.1) Fault-tolerant mode triggering conditions and switching judgment

[0126] Based on the sensor reliability index output in step 2 and the valid observation set in step 3, the system operating mode is determined. Fault-tolerant control mode is triggered when one of the following conditions is met: a sensor exists. Make Or the number of abnormal sensors .

[0127] in: : No. Each sensor at time Credibility; : Indicates an indicator function that takes the value 1 if the condition is true, and 0 otherwise; : Indicates the minimum number of faulty sensors required to trigger the fault-tolerant mode.

[0128] (5.2) Control strategy retention and parameter reconfiguration in fault-tolerant mode

[0129] In fault-tolerant control mode, the minimum functional constraint control strategy constructed in step 4 is maintained, and the control parameters are reconfigured:

[0130]

[0131] in: : Represents the state error weight matrix in fault-tolerant mode; : Represents the adjustment coefficient of the weight matrix; : Represents the set of control input constraints in fault-tolerant mode; : Indicates the control constraint adjustment coefficient.

[0132] (5.3) Sensor state recovery monitoring and reliability reassessment

[0133] During fault-tolerant control mode operation, the system continuously monitors the status of abnormal sensors and reassesses their reliability within a time window.

[0134] in: : No. The average confidence level W of a sensor within a time window: represents the length of the confidence sliding evaluation window; : Length of the confidence sliding evaluation window; : Represents the backtracking index within the window.

[0135] When satisfied At that time, it is considered that the sensor status has been restored.

[0136] in: The sensor recovery determination threshold is met. .

[0137] (5.4) Smooth recovery and switching to normal control mode completed

[0138] When all key sensors involved in the control meet the recovery conditions, the system smoothly switches from fault-tolerant control mode back to normal control mode, and gradually fuses the control inputs:

[0139] in: : Indicates mixed control commands during the switching process; : Represents the smooth switching coefficient that increases over time; : Represents control commands in normal control mode; : Indicates the control command in fault-tolerant control mode.

[0140] satisfy: The output of this step is the system operating mode status and the corresponding control command switching results.

[0141] Step 6) System Real-time Operation and Abnormal Closed-Loop Update Mechanism

[0142] (6.1) Real-time updates of multivariate measurement and reconstructed state

[0143] During vehicle operation, the system uses a uniform sampling period. Real-time acquisition of measurement data from multiple sensor sources, and based on the processing results of steps 1 to 3, a state update operator is used. By combining current observations with historical status information, a real-time updated vehicle status estimate is generated.

[0144] Where: F( ): Indicates a state update operator (such as the Kalman filter update step); : Represents the effective set of observations after integrating real and virtual measurements; : Represents the state estimate at the previous moment.

[0145] (6.2) Closed-loop feedback of abnormal information and self-updating of parameters

[0146] Based on the sensor reliability and mode state information obtained in steps 2 and 5, the anomaly detection threshold and weight parameters are updated in a closed loop:

[0147]

[0148] in: : Indicates the updated threshold for anomaly detection in the next time step; : Indicates the threshold adaptive adjustment amount; : Represents the updated spatiotemporal weight coefficient for the next time step; Δα(k): Represents the adaptive adjustment amount of the weight coefficient.

[0149] (6.3) Suppression of the impact of abnormal conditions on subsequent estimation and control

[0150] To prevent the cumulative impact of historical information from abnormal sensors on subsequent estimation and control, dynamic attenuation processing is applied to the data channels corresponding to abnormal sensors:

[0151] in: : Represents the credibility decay coefficient, satisfying (6.4) System stability determination and operating status output; During real-time operation, the stability of the vehicle status and control results is assessed.

[0152]

[0153] in: : Represents the stability threshold of state changes; : Indicates the stability threshold for changes in control commands.

[0154] When the system continuously meets the above conditions, it is determined that the current operating state is stable, and the vehicle status, control commands and sensor reliability information are output.

[0155] Step 7) Overall Execution Process Management and Termination Condition Determination

[0156] (7.1) Loop scheduling of the overall execution flow of the method

[0157] During vehicle operation, the processing flow consisting of steps 1 to 6 is cyclically scheduled and executed as a complete control cycle, forming a periodic execution sequence.

[0158] in: : Indicates time The method execution flow sequence; "S1"~"S6": represent the functional modules of steps 1 to 6 respectively.

[0159] (7.2) Online monitoring of the effectiveness of method implementation

[0160] During the method's iterative execution, key performance indicators are monitored online to evaluate the effectiveness of the current method execution.

[0161]

[0162] in: Ec(k) represents the magnitude of change in vehicle state estimation; Ec(k) represents the magnitude of change in sensor reliability. (7.3) Persistence of anomalies and determination of system security status A comprehensive assessment of the persistence of sensor anomalies and system operating status is conducted to construct safety status determination conditions: [The following is a partial sentence and requires more context for accurate translation] (Continuous decision window length), so that for all All meet And the number of abnormal sensors in: : Indicates the length of the continuous safety decision window; : Indicates the start time of the window; : Indicates the maximum allowed number of abnormal sensors threshold.

[0163] (7.4) Method termination conditions and running status output

[0164] The current method execution cycle or fault tolerance control flow will terminate when any of the following conditions are met: ( (the termination time of method execution), or (The vehicle's state continues to violate the minimum functional physical constraints).

[0165] When the method terminates or the cycle is completed, the system outputs the vehicle operating status, control command sequence, sensor reliability evolution results, and fault-tolerant control execution flag, ultimately forming an integrated closed-loop mechanism of "data fusion → state estimation → fault-tolerant control".

Claims

1. A vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors, characterized in that: Step 1) Unified modeling and physical consistency constraint construction of multivariable measurement sensors: acquire multivariable measurement data from multiple source sensors of the vehicle, and perform time alignment processing on the measurement data of different sensors to construct a synchronized multivariable measurement set; at the same time, define the key physical state vector of the vehicle that is independent of the specific sensor type, and establish a unified state space model of the vehicle state. Step 2) Multivariate measurement consistency analysis and anomaly determination based on spatiotemporal redundancy: Based on the unified state space model, observation models between multi-source sensor measurements and vehicle state are established respectively, and sensor anomaly deviation terms are introduced; at the same time, a vehicle dynamics state transition model is established, and a set of physical consistency constraints that the vehicle state must satisfy is constructed. Step 3) State reconstruction and virtual sensor generation based on credibility weighting: Within a preset time window, based on the temporal and spatial redundancy characteristics of multi-source sensor measurements, spatiotemporal consistency analysis is performed on the measurement results of each sensor, the spatiotemporal consistency metric corresponding to each sensor is calculated, and the credibility index of each sensor is obtained by combining the physical consistency constraint set. Step 4) Construction of real-time fault-tolerant control strategy for vehicle minimum functional constraints: Based on the credibility index of each sensor, the multi-source sensor measurements are weighted according to credibility to construct the reconstruction estimation result of the vehicle's key state; when the credibility of any sensor is lower than a preset threshold, the corresponding virtual sensor measurement is generated based on the reconstruction estimation result, and the virtual measurement replaces the measured data of the abnormal sensor to form an effective observation set. Step 5) Fault-tolerant mode switching and normal control recovery mechanism: Based on the reconstructed state and effective observation set, construct a set of state constraints that meet the minimum functional operation requirements of the vehicle, and solve the fault-tolerant control command of the vehicle under this set of state constraints, so that the vehicle can maintain basic driving and safety control functions in the event of sensor abnormality or failure. Step 6) Real-time operation and abnormal closed-loop update mechanism of the system: Based on the changes in sensor reliability index and vehicle operating status, switch between normal control mode and fault-tolerant control mode, and reconfigure control parameters in fault-tolerant control mode. At the same time, abnormal sensors are continuously monitored. When the sensor reliability recovers to the preset recovery threshold, smoothly switch back to normal control mode. Step 7) Overall execution process management and termination condition determination: The processing flow consisting of steps 1) to 6) is executed cyclically with a preset sampling period, and the stability of vehicle status and the persistence of sensor anomalies are determined online. When the preset termination condition is met, the vehicle operating status, control commands and sensor reliability information are output, and the method execution is completed.

2. The vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors according to claim 1, characterized in that, The specific method in step 1) is as follows: Given the possibility of sensor malfunction or failure, a unified state description model independent of specific sensor types is constructed, and vehicle physical consistency constraints are introduced in advance to provide a basic model and judgment basis for subsequent anomaly identification, state reconstruction and fault-tolerant control. During vehicle operation, data is collected from multiple onboard sensors at discrete moments. The original measurement data are used to form a multivariate measurement set: in, Indicates the number of sensors. Indicates the first The observation vector obtained by each sensor at its own sampling time; Because different sensors have different sampling frequencies and timestamps, the measurement data from each sensor are aligned to a unified time axis and mapped to a unified system sampling time. : in, To ensure a unified sampling period for the system, For the first The mapping relationship between the original timestamps of each sensor and the unified time grid; After time alignment is completed, a set of synchronized measurements is obtained for subsequent modeling: Based on vehicle kinematics and control requirements, the vehicle is defined at time... The key physical state vector is: in, This represents the vehicle's position coordinates in three-dimensional space. Represents the vehicle speed component. Represents the vehicle acceleration component; The key physical state vector serves as a unified expression of the vehicle's operating state, independent of the specific sensor type and number, and is used to describe the core physical characteristics of the vehicle in terms of spatial position, motion state, and dynamic changes. Regarding the first A unified observation model is established between the measurement data of each sensor and the vehicle status: in , For the first Measurement mapping function for each sensor, To measure the noise vector, This is the corresponding measurement noise covariance matrix; To describe the sensor's measurement characteristics under abnormal, drifting, or disturbed conditions, anomaly terms are introduced into the observation model: in, Indicates the first Each sensor at time The abnormal deviation term is used to characterize the systematic errors or abnormal disturbances that may exist in the sensor output. Based on the vehicle's kinematics or dynamics, construct a discrete-time transition model of the vehicle's state: Process noise , This represents the vehicle state transition function. This is the vehicle control input vector. This is the process noise vector. The process noise covariance matrix; Based on vehicle physical accessibility and safe operation requirements, a set of physical consistency constraints that the vehicle state must satisfy is constructed: Define the velocity and acceleration constraints as follows: in, and These represent the maximum permissible speed and maximum acceleration threshold of the vehicle under the current operating conditions, respectively. After completing the above processing, a unified modeling result is output, including the vehicle state space model, synchronous measurement data, observation model parameters, state transition model, and physical consistency constraint set, which are used for subsequent spatiotemporal redundancy analysis and sensor anomaly determination. During system operation, measurement data of the same physical state from multiple sensors are collected simultaneously, forming a multi-source measurement set corresponding to each moment. in, This represents the set of all sensor measurements collected at time tn. j Let represent the measurement value of the j-th sensor at time tn; M is the number of sensors; Each sensor has a unique number j during the data acquisition process, j = 1, 2, … , M, which is used for subsequent error analysis and consistency determination; Based on historical measurement samples, experimental calibration results, or sensor technical parameters, determine the random noise amplitude range and communication delay range of each sensor. Noise amplitude range is defined as , where ε + This indicates the maximum deviation of the sensor's measurement noise; the communication delay range is defined as... , where τ - and τ + These represent the minimum delay and the maximum delay, respectively. By determining the error distribution characteristics of each sensor's measurement output using the above parameters, an error boundary model that comprehensively reflects noise and delay uncertainties is established, providing a quantitative basis for subsequent boundary correction and fusion analysis.

3. The vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors according to claim 1, characterized in that, The specific method in step 2) is as follows: Based on the unified state model, observation model, and physical consistency constraints constructed in step 1), consistency analysis is performed using the spatiotemporal redundancy relationship of multi-source sensor measurements to achieve rapid determination of sensor abnormal states; for the first... Each sensor constructs its measurement sequence within a continuous time window: in, The length of the time window; Based on predicted state Calculate the corresponding predicted observations: Constructing the time residual sequence: And calculate the time consistency metric: in, For the first The measurement noise covariance matrix of each sensor. Used to measure the stability and consistency of the sensor's measurements over time; For a set of sensors capable of observing the same vehicle state variable Construct a cross-sensor spatial consistency analysis model; for any two sensors The observed difference is defined as: Constructing a spatial consistency index based on observation noise characteristics: Further on the first A comprehensive evaluation of the spatial consistency of each sensor is conducted. in, Indicates the first The degree of spatial consistency of each sensor relative to other redundant sensors; By combining time consistency metrics and spatial consistency metrics, the first... Spatiotemporal consistency measure of individual sensors: in, The consistency weighting coefficients for the time and space dimensions are used to adjust the importance of the two types of redundant information according to the system's operating conditions. Introducing the physical consistency constraint from step 1), for the first... The state components are estimated by a single sensor. Perform constraint checks: When constraints are violated, a physical default indicator is constructed: in, This is used to reflect the degree of deviation between the sensor's measurement results and the vehicle's physically feasible domain; By integrating spatiotemporal consistency measures and physical consistency indices, the first... Anomaly detection function for each sensor: in, This is the penalty coefficient for physical consistency. Calculate sensor reliability based on anomaly detection function: in, As a confidence threshold, when the above conditions are met, the first [condition] is determined. Each sensor at time It is in an abnormal state; The final output includes the anomaly detection result and corresponding reliability index for each sensor. It serves as the direct input for sensor weight adjustment, state reconstruction, and virtual sensor generation in step 3).

4. The vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors according to claim 1, characterized in that, The specific method in step 3) is as follows: In step 2), the time of each sensor is obtained. Credibility index Then, the weights of the multi-source sensors involved in state reconstruction are normalized to construct a confidence-weighted coefficient: in, The number of sensors currently involved in the estimation. Indicates the first The relative contribution weights of each sensor to the vehicle state estimation at the current moment, satisfying: Based on the unified state-space model and observation model constructed in step 1), the key states of the vehicle are reconstructed and estimated using confidence-weighted multi-source measurements: in, For vehicles at any time The reconstructed state estimation vector, For the time-aligned first Each sensor measurement vector, The corresponding observation mapping matrix is ​​calculated as follows: in, This is the linearized matrix of the observation function at the current estimation point. For the first The measurement noise covariance matrix of each sensor; When a sensor is determined to be abnormal in step 2), that is, its confidence level satisfies: ,in, To preset a confidence threshold, the vehicle state obtained from the reconstruction is then... The virtual measurement output of the sensor is generated through the corresponding observation function: in, Used to replace actual measured data in subsequent control and judgment during sensor malfunction or failure; Based on the sensor's reliability status, the real and virtual measurements are adaptively switched and fused to form the final usable observations: in, The confidence-based switching coefficient is defined as follows: Through the above processing, when the sensor is in normal condition, its actual measured data is used directly; when the sensor malfunctions or fails, the output of a virtual sensor is used instead, ensuring the continuity and consistency of vehicle state information under abnormal operating conditions; finally, the vehicle state reconstruction estimation result is output. and the fused effective observation set This serves as the input for determining the vehicle's minimum functional constraints and making fault-tolerant control decisions in subsequent steps.

5. The vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors according to claim 1, characterized in that, The specific method in step 4) is as follows: Based on the vehicle reconfiguration state estimate and effective observation set output in step 3), a constraint system and control objective are constructed around the minimum functional operation requirements of the vehicle, and the optimal fault-tolerant control command is solved to ensure safe driving under abnormal conditions. Based on the basic driving and safety requirements that the vehicle must maintain even under sensor malfunctions, a set of minimum functional operating state constraints is determined to limit the feasible range of vehicle states during fault-tolerant control: in, For the set of physical consistency constraint functions constructed in step 1), the typical form of the minimum functional constraint is further given: in, To ensure a safe speed limit, Let yaw rate be the vehicle's angular velocity. The yaw stability threshold. The above thresholds are preset as a safe acceleration limit based on vehicle dynamics characteristics and safety strategies; The vehicle reconfiguration state estimate output in step 3) Based on this, a fault-tolerant control objective function oriented towards minimum functional constraints is constructed: in, This is the reference state vector in the minimum functional operating mode. This is a state error weighting matrix used to adjust the importance of different state variables in the control objective; Reference status Based on the current driving conditions, priority is given to ensuring the vehicle's longitudinal stability, lateral controllability, and attitude safety, without pursuing high-precision trajectory tracking. Under the constraints of the vehicle dynamics model, the control input is optimized and solved: in, This is the vehicle control input vector. This is the set of control inputs allowed in the minimum functional mode, used to limit the range of changes in steering angle, driving force, or braking force. The optimal fault-tolerant control command obtained under constraints; The obtained optimal fault-tolerant control command It acts on the vehicle's actuators and monitors the vehicle's operating status in real time. When the vehicle state continuously satisfies the minimum functional constraint set During this period, the system maintains fault-tolerant control mode; when the sensor reliability recovers to above the threshold, it provides conditions for exiting fault-tolerant mode and resuming normal control. The output of this step is a sequence of real-time fault-tolerant control instructions that meet the minimum functional constraints. This is a direct result of control measures to maintain safe operation of the vehicle in the event of sensor malfunction or failure.

6. The vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors according to claim 1, characterized in that, The specific method in step 5) is as follows: Based on sensor reliability status and vehicle operating conditions, dynamic switching between normal control mode and fault-tolerant control mode is achieved. Based on the sensor reliability index output in step 2) and the effective observation set in step 3), the system operating mode is determined, and the fault-tolerant control mode is triggered when one of the following conditions is met: in, For the first Each sensor at time Credibility, As a credibility threshold, For indicator functions, The minimum number of abnormal sensors required to trigger the fault-tolerant mode; when the above conditions are met, the system switches from normal control mode to fault-tolerant control mode. In fault-tolerant control mode, the minimum functional constraint control strategy constructed in step 4) is maintained, and the control parameters are reconfigured: in, This is the state error weight matrix in fault-tolerant mode. This is the set of control input constraints in fault-tolerant mode. and The adjustment coefficient is used to reduce the system's reliance on high-precision tracking performance and prioritize vehicle stability and safety. During fault-tolerant control mode operation, the system continuously monitors the status of abnormal sensors and reassesses their reliability within a time window: in, The length of the confidence sliding evaluation window, For the first The average confidence level of a sensor within a time window is used to determine whether the sensor state has recovered, provided that the following conditions are met: in, To restore the judgment threshold, and satisfy ; When all key sensors involved in control meet the recovery conditions, the system smoothly switches from fault-tolerant control mode back to normal control mode; to avoid abrupt changes in control commands, the control inputs are progressively fused. in, These are control commands in normal control mode. These are control commands for fault-tolerant control mode. For a smooth switching coefficient that increases over time, the following condition must be met: when When switching from fault-tolerant control mode to normal control mode, the system outputs the system operating mode status and the corresponding control command switching result, ensuring the safe and continuous operation of the vehicle after the sensor anomaly is eliminated.

7. The vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors according to claim 1, characterized in that, The specific method in step 6) is as follows: The system updates vehicle status and system parameters cyclically with a uniform sampling period, and suppresses anomalies through closed-loop feedback and anomaly suppression. During vehicle operation, the system updates the vehicle status and system parameters cyclically with a uniform sampling period T. s Real-time acquisition of measurement data from multiple sensors, and based on the processing results from steps 1) to 3), the state update operator F( By combining current observations with historical status information, a real-time updated vehicle status estimate is generated: in, To create an effective set of observations that integrates real and virtual measurements, This represents a state update operator used to integrate current observations and historical state information to achieve continuous estimation of the vehicle's critical states. Based on the sensor reliability and mode state information obtained in steps 2) and 5), the anomaly detection threshold and weight parameters are updated in a closed loop. Where Δc(k) and Δα(k) are adaptive adjustment values ​​calculated based on the current operational stability and the duration of the anomaly, respectively; to prevent the historical information from the anomaly sensor from having a cumulative impact on subsequent estimation and control, the data channel corresponding to the anomaly sensor is dynamically attenuated. in, This is the reliability attenuation coefficient, used to reduce the impact of abnormal sensors on state estimation and control decisions during the duration of the anomaly; To reduce the impact of abnormal sensors on state estimation and control decisions during the period of anomaly, and to determine the stability of vehicle state and control results during real-time operation: in, and These are the state change threshold and the control change threshold, respectively. When the system continuously meets the above conditions, it is determined that the current operating state is stable, and the vehicle state, control command and sensor reliability information are output as inputs for the next control cycle and the upper-level decision module.

8. The vehicle fault-tolerant control method under abnormal conditions of multivariable measurement sensors according to claim 1, characterized in that, The specific method in step 7) is as follows: During vehicle operation, the processing flow consisting of steps 1) to 6) is cyclically scheduled and executed as a complete control cycle. The system performs this cyclically in each sampling cycle. The system sequentially completes multivariate measurement acquisition, spatiotemporal redundancy and consistency analysis, state reconstruction and virtual sensor generation, minimum function fault-tolerant control, mode switching and closed-loop update processing, forming a periodic execution sequence. in, Indicates at time The complete execution process ensures that all functional modules coordinate in an orderly manner in time and operate in a closed loop in logic; During the method's iterative execution, key performance indicators are monitored online to evaluate the effectiveness of the current method execution. in, This indicates the magnitude of change in the estimated vehicle condition. This indicates the change in sensor reliability; when the above indicators remain within the preset threshold range, the method is considered stable and effective. A comprehensive assessment of the persistence of sensor anomalies and system operating status is conducted to construct safety status determination conditions: in, To continuously determine the window length, The maximum allowed number of abnormal sensors is set as a threshold. When the vehicle status consistently meets the physical consistency constraint within a continuous window and the number of abnormal sensors does not exceed the threshold, the system is determined to be in a safe operating state. The current method execution cycle or fault tolerance control flow will terminate when any of the following conditions are met. ,in The termination time of the method execution; or This indicates that the vehicle's state continuously violates the minimum functional physical constraints. When the method terminates or the cycle is completed, the system outputs the vehicle's operating status, control command sequence, sensor reliability evolution results, and fault-tolerant control execution flags, which are used by the vehicle's upper-level decision-making system for recording, analysis, or triggering manual takeover and other subsequent processing.

Images