An Adaptive Forgetting Extended Kalman Filter Tire Pressure Monitoring Method and Device

Abstract

This invention discloses an adaptive forgetting extended Kalman filter tire pressure monitoring method, belonging to the field of vehicle indirect tire pressure monitoring systems. By constructing a seven-DOF suspension dynamics model including tire radial stiffness, the radial stiffness of the four wheels is extended into state variables to form an 18-dimensional state vector. An 8-dimensional output vector is defined to establish a nonlinear state-space equation. A genetic algorithm combined with CarSim simulation data is used to identify model parameters to obtain an accurate dynamic model. Based on these model parameters, a state observer is designed, and a double forgetting factor is introduced to optimize its dynamic tracking capability and steady-state estimation accuracy. Tire pressure data is monitored in real time, tire pressure is estimated through the state observer, and a blowout detection algorithm is constructed based on dimension-unified innovation squares to achieve rapid blowout detection and early warning. This invention significantly improves the accuracy of tire pressure estimation and the real-time performance of blowout detection.

Description

Technical Field

[0001] This invention relates to the field of indirect tire pressure monitoring systems for vehicles, and in particular to an adaptive forgetting extended Kalman filter tire pressure monitoring method, apparatus, device, and storage medium. Background Technology

[0002] As the only part of a vehicle in contact with the ground, tire inflation pressure is a key parameter for ensuring vehicle safety, economy, and performance. Maintaining proper tire pressure is crucial for overall vehicle performance. From an economic perspective, insufficient tire pressure significantly increases rolling resistance, leading to increased fuel consumption. Studies show that for every 0.2 bar drop in tire pressure below the rated value, fuel economy is negatively impacted, and approximately 20% of tire wear life is lost. From a safety perspective, the dangers of insufficient tire pressure are even more severe: First, increased tire sidewall deformation lowers the critical speed for standing wave formation, making the tire more susceptible to blowouts at high speeds due to heat buildup or structural fatigue, potentially causing serious traffic accidents. Second, altered tire contact characteristics affect vehicle handling stability, braking performance, and steering response, weakening the driver's control. Especially during a blowout, the sudden change in tire pressure causes a sharp deterioration in vehicle dynamics, posing an extreme threat to driving safety.

[0003] Currently, tire pressure monitoring (TPM) mainly falls into two technical categories: direct measurement and indirect estimation. While direct monitoring offers higher accuracy, it suffers from inherent limitations such as high sensor cost, the need for additional power and maintenance, and susceptibility to wireless signal interference. Traditional indirect monitoring methods, on the other hand, rely heavily on limited signals like wheel speed, making them ill-suited for complex driving conditions and changes in system parameters. They struggle to accurately identify slow tire pressure leaks in the early stages and cannot reliably capture the dynamic changes in tire pressure during a blowout. Consequently, they fail to provide timely and accurate information on abnormal tire pressure status for subsequent vehicle active safety control systems (such as self-stabilization control after a blowout).

[0004] Therefore, there is an urgent need for a highly reliable monitoring method that does not rely on direct pressure sensors, can accurately estimate tire pressure in real time, and can sensitively identify sudden changes in tire blowout. This method would compensate for the shortcomings of existing technologies and provide key technical perception support for improving vehicle active safety and building a tire blowout emergency safety barrier. Summary of the Invention

[0005] The present invention aims to at least partially solve one of the technical problems in the related art.

[0006] To address this, this invention proposes an adaptive forgetting extended Kalman filter tire pressure monitoring method. It constructs a nonlinear state-space equation for the suspension, incorporating tire radial stiffness, and uses CarSim simulation data and a genetic algorithm for parameter identification to improve model accuracy. Furthermore, a state observer is designed, incorporating a dual forgetting factor to optimize its dynamic tracking capability and steady-state estimation accuracy, achieving an adaptive balance in observation performance. Based on this, tire pressure is estimated in real-time using this observer, and a blowout detection algorithm is constructed based on dimension-unified innovation squares. Ultimately, this achieves continuous and accurate tire pressure monitoring and rapid blowout warning, effectively improving the reliability of vehicle active safety under complex operating conditions.

[0007] Another objective of this invention is to provide an adaptive forgetting extended Kalman filter tire pressure monitoring device.

[0008] The third objective of this invention is to provide a computer device.

[0009] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0010] To achieve the above objectives, this invention proposes an adaptive forgetting extended Kalman filter tire pressure monitoring method, comprising:

[0011] S1. A seven-degree-of-freedom suspension dynamics model including the radial stiffness of the four tires is constructed. A 14-dimensional state vector is selected, which includes the vertical displacement of the vehicle's center of gravity, pitch angle, roll angle, unsprung mass displacement of the four wheels, and their first derivatives. The radial stiffness of the four tires is extended into system state variables to form an 18-dimensional state vector. An 8-dimensional output vector including the sprung mass acceleration and unsprung mass acceleration of the four wheels is defined. A nonlinear state-space equation is established based on the seven-degree-of-freedom suspension dynamics model. The inherent parameters of the seven-degree-of-freedom suspension dynamics model are identified using a genetic algorithm with CarSim simulation data as a reference, so as to construct more accurate dynamic model parameters.

[0012] S2, Design a state observer based on the parameters of the dynamic model, and introduce two forgetting factors to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer;

[0013] S3 monitors tire pressure data in real time and estimates tire pressure results through a state observer. Based on the tire pressure results and the state observer, it generates a dimensionally unified information squared burst identification algorithm to achieve rapid detection and early warning of bursts.

[0014] An adaptive forgetting extended Kalman filter tire pressure monitoring method according to an embodiment of the present invention may also have the following additional technical features:

[0015] In one embodiment of the present invention, a nonlinear state-space equation is established based on the seven-degree-of-freedom suspension dynamics model, including:

[0016] S11. Based on an 18-dimensional state vector and an 8-dimensional output vector, a nonlinear state-space model containing state transition equations and measurement equations is established. The state transition equations describe the relationship between the state vector and time, while the measurement equations describe the relationship between the output vector and the state vector and input vector.

[0017] In one embodiment of the present invention, two forgetting factors are introduced to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer, including:

[0018] S21, using the forgetting factor Reduce the observer's reliance on historical observation data:

[0019] ;

[0020] in, A constant forgetting factor. Here is the state transition matrix. The process noise covariance matrix is... The prior state estimation error covariance matrix is... The posterior state estimation error covariance matrix is... for transpose;

[0021] S22, by introducing an improved Sage-Husa adaptive filter, dynamically corrects the measurement noise covariance matrix. :

[0022] ;

[0023] in, These are weighting coefficients. A constant forgetting factor. , For the innovation vector, , Estimate the measurement noise covariance matrix and mean at the current moment. For prior state estimation, The original observation residuals, For the estimation of the measurement noise covariance matrix at the next time step, Let be the current information vector. For the new information vector transpose, To estimate the mean vector of the measurement noise at the next time step, This represents the actual sensor measurement at the current moment. For the observation model function, For the estimation of the prior state at the current moment, For the system input at the current moment, Forgetting factor The kth power.

[0024] In one embodiment of the present invention, a blowout identification algorithm based on tire pressure results and a state observer generating dimensionally unified innovation squares is used to achieve rapid detection and early warning of tire blowouts, including:

[0025] S31, calculate the square of dimensionless innovation;

[0026] S32, Set the tire blowout detection threshold to 200, when the calculated value... When the threshold is exceeded, a tire blowout is determined to have occurred and a warning signal is triggered.

[0027] In one embodiment of the present invention, it further includes:

[0028] S4, Dynamic compensation steps based on vehicle shifting and steering conditions:

[0029] S41, during vehicle braking, deceleration, and steering, adjust the input variables of the state transition equation according to the real-time vehicle speed and steering angle;

[0030] S42 converts the dynamically estimated tire radial stiffness into a tire pressure estimate to improve the accuracy of tire pressure estimation under variable speed steering conditions.

[0031] To achieve the above objectives, another aspect of the present invention proposes an adaptive forgetting extended Kalman filter tire pressure monitoring device, comprising:

[0032] The model building module is used to construct a seven-degree-of-freedom suspension dynamics model that includes the radial stiffness of the four tires. A 14-dimensional state vector is selected, which includes the vertical displacement of the vehicle's center of gravity, pitch angle, roll angle, unsprung mass displacement of the four wheels, and their first derivatives. The radial stiffness of the four tires is augmented into system state variables to form an 18-dimensional state vector. An 8-dimensional output vector is defined, which includes the sprung mass acceleration and unsprung mass acceleration of the four wheels. A nonlinear state-space equation is established based on the seven-degree-of-freedom suspension dynamics model. The inherent parameters of the seven-degree-of-freedom suspension dynamics model are identified using a genetic algorithm with CarSim simulation data as a reference, so as to construct more accurate dynamic model parameters.

[0033] The state observer design and optimization module is used to design a state observer based on the parameters of the dynamic model, and introduce two forgetting factors to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer.

[0034] The tire blowout condition identification module is used to monitor tire pressure data in real time and estimate the tire pressure result through a state observer. Based on the tire pressure result and the state observer, a dimension-unified innovation squared tire blowout identification algorithm is generated to achieve rapid detection and early warning of tire blowout.

[0035] In one embodiment of the present invention, it further includes:

[0036] The dynamic compensation module is used for dynamic compensation steps based on vehicle shifting and steering conditions: during vehicle braking, deceleration, and steering, the input variables of the state transition equation are adjusted according to the real-time vehicle speed and steering angle; the dynamically estimated tire radial stiffness is converted into a tire pressure estimate to improve the tire pressure estimation accuracy under shifting and steering conditions.

[0037] This invention discloses an adaptive forgetting extended Kalman filter tire pressure monitoring method and device. By constructing a high-precision seven-DOF suspension-tire dynamics model and integrating genetic algorithm parameter identification with a dual-forgetting factor adaptive observer design, it effectively solves the problems of large tire pressure estimation deviation and high warning delay caused by model mismatch and noise interference in traditional methods under complex dynamic conditions. It achieves integrated monitoring from high-fidelity dynamic modeling and adaptive state estimation to rapid tire blowout identification based on innovation statistics, significantly improving the accuracy, real-time performance, and reliability of tire pressure monitoring under different driving conditions, and enhancing the system's robustness and applicability in real-world complex driving environments.

[0038] To achieve the above objectives, a third aspect of this application provides a computer device comprising a processor and a memory; wherein the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory, for implementing an adaptive forgetting extended Kalman filter tire pressure monitoring method as described in the first aspect embodiment.

[0039] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an adaptive forgetting extended Kalman filter tire pressure monitoring method as described in the first aspect embodiment.

[0040] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0041] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0042] Figure 1This is a flowchart of an adaptive forgetting extended Kalman filter tire pressure monitoring method according to an embodiment of the present invention;

[0043] Figure 2 This is a roadmap of a vehicle-wide indirect tire pressure fast estimation and blowout identification algorithm for another adaptive forgetting extended Kalman filter tire pressure monitoring method according to an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of a seven-degree-of-freedom vehicle suspension model of another adaptive forgetting extended Kalman filter tire pressure monitoring method according to an embodiment of the present invention;

[0045] Figure 4(a) illustrates how the proposed FEKF and AFEKF algorithms quickly and stably converge to the true value after a sudden change in tire pressure, with convergence times of 0.35 s and 0.62 s, respectively. In contrast, the conventional EKF algorithm cannot dynamically track tire pressure changes. Figure 4(b) shows the time-domain statistics of the dimensionless innovation square. After a sudden change in tire pressure within 20 s, the dimensionless innovation square surges and produces a significant peak, exceeding the set threshold, which indicates a tire blowout and issues a warning signal. Figure 4(c) illustrates the tire pressure estimation results of the four tires using the FEKF algorithm under the condition of a front left tire blowout. Figure 4(d) illustrates the tire pressure estimation results of the four tires using the AFEKF algorithm under the condition of a front left tire blowout.

[0046] Figure 5(a) shows the FEKF results; Figure 5(b) shows that the peak-to-peak value of the AFEKF tire pressure estimation has been optimized and the fluctuation amplitude has been significantly suppressed.

[0047] Figure 6(a) shows the estimated front left tire pressure under variable speed conditions; Figure 6(b) shows the lateral displacement trajectory of the vehicle; Figure 6(c) shows the dimensionless squared result of the variable speed condition; Figure 6(d) shows the driving speed curve.

[0048] Figure 7(a) illustrates that AFEKF still maintains high tire pressure estimation accuracy on ISO-B, while the FEKF algorithm's tire pressure estimation oscillates severely, and the error is further amplified; Figure 7(b) illustrates that AFEKF still maintains high tire pressure estimation accuracy on the actual road surface, while the FEKF algorithm's tire pressure estimation oscillates severely, and the error is further amplified.

[0049] Figure 8 This is a schematic diagram of the structure of an adaptive forgetting extended Kalman filter tire pressure monitoring device according to an embodiment of the present invention;

[0050] Figure 9 It is a computer device according to an embodiment of the present invention. Detailed Implementation

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0052] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0053] The following description, with reference to the accompanying drawings, describes an adaptive forgetting extended Kalman filter tire pressure monitoring method, apparatus, device, and storage medium according to embodiments of the present invention.

[0054] The core idea of ​​this invention is to construct a seven-degree-of-freedom high-fidelity suspension-tire coupled dynamics model that includes the radial stiffness of four tires, and to extend the radial stiffness of the tires to the system state variables to establish a nonlinear state-space equation. At the same time, a genetic algorithm combined with CarSim simulation data is used to accurately identify the inherent parameters of the model. Building upon this foundation, an extended Kalman filter state observer is designed based on the identified model parameters, and a dual forgetting factor collaborative optimization mechanism is introduced. The first forgetting factor weakens the dependence on historical observation data to enhance dynamic tracking capabilities, while the second forgetting factor dynamically corrects the statistical characteristics of measurement noise through an improved Sage-Husa adaptive filter, thus achieving an adaptive balance between rapid response and steady-state accuracy. Furthermore, this observer is used to estimate tire radial stiffness in real time and convert it into tire pressure values, constructing a blowout identification algorithm based on dimensionally unified innovation squares to achieve rapid detection and early warning. To cope with complex driving conditions, this invention also includes a dynamic compensation step, dynamically adjusting the state transition equation input during vehicle speed changes or steering and combining real-time vehicle speed and steering angle for error correction. Ultimately, the traditional tire pressure monitoring system, which relies on direct measurement, is transformed into an integrated intelligent monitoring system that combines a high-precision model, adaptive estimation, statistical decision-making, and dynamic compensation. This significantly improves the accuracy of tire pressure estimation, the real-time performance of blowout warnings, and the overall robustness and reliability of the system under dynamic and interference environments.

[0055] Example 1

[0056] To achieve the above invention, embodiments of the present invention provide an adaptive forgetting extended Kalman filter tire pressure monitoring method, such as... Figure 1 As shown, it includes:

[0057] S1. Construct a seven-degree-of-freedom suspension dynamics model that includes the radial stiffness of the four tires. Select a 14-dimensional state vector that includes the vertical displacement of the vehicle's center of gravity, pitch angle, roll angle, unsprung mass displacement of the four wheels, and their first derivatives. Extend the radial stiffness of the four tires into system state variables to form an 18-dimensional state vector. Define an 8-dimensional output vector that includes the sprung mass acceleration and unsprung mass acceleration of the four wheels. Based on the seven-degree-of-freedom suspension dynamics model, establish a nonlinear state-space equation. Use a genetic algorithm and CarSim simulation data as a reference to identify the inherent parameters of the seven-degree-of-freedom suspension dynamics model to construct more accurate dynamics model parameters.

[0058] Specifically, by constructing a high-fidelity vehicle suspension-tire coupled dynamic model, the key mechanical parameter of tire radial stiffness, which is strongly correlated with tire pressure, is extended to represent the internal state of the system. This establishes a nonlinear state-space description framework that reflects the impact of tire pressure changes on the dynamic response of the suspension system. This principle is based on the monotonic mapping relationship between tire pressure and tire radial stiffness. By observing the measurable outputs of the suspension system (such as sprung / unsprung mass acceleration), the tire radial stiffness, which cannot be directly measured, is inferred using state estimation theory, thereby indirectly obtaining tire pressure information.

[0059] Specifically, a seven-DOF lumped-parameter dynamic model is first established, encompassing the vertical, pitch, and roll motions of the vehicle body, as well as the unsprung mass motion of the four wheels. Fourteen basic states are selected, including the vertical displacement of the vehicle's center of gravity, pitch angle, roll angle, and the displacements of the unsprung mass of the four wheels, along with their first derivatives. The radial stiffness of the four tires is then added as an estimated parameter to the state vector, forming an 18-dimensional system state. The model's output is defined as the accelerations of the sprung and unsprung masses of the four wheels, totaling eight physical quantities that can be directly measured by onboard acceleration sensors. Based on the Newton-Euler equations, the state transition equations and measurement equations of the system are derived, forming a complete nonlinear state-space expression. This provides a precise mechanistic model foundation for the subsequent design of the state observer.

[0060] Furthermore, to improve the matching accuracy between the model and the actual vehicle dynamic characteristics, a genetic algorithm is used to identify the inherent parameters of the suspension model (such as sprung mass, moment of inertia, suspension stiffness, and damping coefficient) offline. The sprung / unsprung mass acceleration data generated by the high-fidelity vehicle dynamics software CarSim under typical conditions (such as ISO-A road surface and constant speed of 80 km / h) are used as a reference benchmark. A cost function is constructed using the sum of the multi-channel root mean square errors between the model output response and the reference data. Through iterative optimization using the genetic algorithm, the optimal parameter set that minimizes the model error is obtained, thereby ensuring the high accuracy and reliability of the dynamic model parameters.

[0061] Specifically, this high-precision modeling and parameter identification method lays a solid foundation for subsequent real-time tire pressure estimation. This method does not rely on direct tire pressure sensors or wheel speed signals, but instead utilizes information from suspension acceleration sensors widely found in modern vehicles, offering advantages such as low cost and ease of deployment. Its application scenarios broadly cover various real-world driving conditions, including high-speed straight-line driving, gear shifting, and steering, and it provides accurate model support for handling extreme tire blowout conditions where single or even multiple wheels experience varying degrees of pressure loss simultaneously.

[0062] Specifically, the accurate dynamic model and parameter set constructed through the above steps fundamentally improve the model fidelity of the state observer design. Compared with methods based on simplified or empirical models, this invention effectively reduces estimation errors caused by model mismatch, providing a key prerequisite for achieving high-precision and robust real-time estimation of tire radial stiffness and tire pressure, and ensuring the overall performance of subsequent adaptive filtering algorithms and tire blowout detection functions from the source. Simulation results show that the estimation system based on this accurate model can converge quickly after a sudden change in tire pressure and maintain excellent estimation accuracy and stability under various complex working conditions.

[0063] Furthermore, S1 includes:

[0064] S11. Based on an 18-dimensional state vector and an 8-dimensional output vector, a nonlinear state-space model containing state transition equations and measurement equations is established. The state transition equations describe the relationship between the state vector and time, while the measurement equations describe the relationship between the output vector and the state vector and input vector.

[0065] Specifically, by defining an 18-dimensional state vector encompassing vehicle body motion, wheel motion, and key characteristic parameters (tire radial stiffness), and an 8-dimensional output vector composed of directly or indirectly measurable vehicle response signals, the foundation for model-based estimation theory is laid. Within this vector space framework, the state transition equations, strictly adhering to Newton-Euler's laws, characterize the evolution of the state vector (including tire radial stiffness) under the influence of internal system dynamics and external inputs in the form of a system of differential equations. The measurement equations establish the mapping relationship between the system's internal hidden states and external observable outputs, transforming theoretical state quantities into physical quantities that can be compared with actual sensor data.

[0066] Specifically, the state transition equation (f) is organized and discretized based on the seven-degree-of-freedom suspension dynamics differential equation, and its specific form is a multidimensional nonlinear function, with the input being the state estimate of the previous time step (f). ) and current control input ( The output is the predicted value of the current state. The measurement equation (h), based on the vehicle body geometry and mechanical equilibrium principles, converts the vehicle body acceleration, pitch acceleration, roll acceleration, and unsprung mass acceleration calculated from the tire dynamics equations in the state vector into sprung and unsprung mass accelerations corresponding to the four wheel positions, resulting in eight channels of predicted observations. These two equations together constitute a complete nonlinear system model for iterative calculations using the extended Kalman filter.

[0067] Furthermore, the key parameters involved in this model are clearly defined: the state space is 18-dimensional, specifically including 1-dimensional vertical displacement of the vehicle body, pitch angle, roll angle, and 3-dimensional velocity; 4-dimensional displacement and 4-dimensional velocity of the four unsprung masses; and 4-dimensional radial stiffness of the four tires. The output space is 8-dimensional, namely the vertical acceleration of the sprung mass at the four wheels and the vertical acceleration of the four unsprung masses. The inherent physical parameters in the model, such as vehicle body mass (…), are also specified. ), moment of inertia ( ), suspension stiffness and damping coefficient ( and unsprung mass ( All of these are identified and optimized using a genetic algorithm that takes CarSim high-fidelity simulation data as a reference, ensuring that their values ​​accurately reflect the characteristics of actual vehicles.

[0068] Specifically, the application scenarios cover the vehicle's state under various road surface excitations (such as random road surfaces and pulsed road surfaces) and multiple driving conditions (including constant speed straight line, acceleration, braking, and steering). In practical applications, the model's input ( The system will receive real-time vehicle bus signals (such as vehicle speed and steering angle) or road disturbance information provided by the anti-road system, and the output will be compared with the measured data of the acceleration sensor installed on the suspension to drive the subsequent adaptive filtering algorithm.

[0069] Specifically, by establishing this high-dimensional, high-fidelity nonlinear state-space model and augmenting it with tire radial stiffness as a state variable, a precise physical information foundation is provided for model-based state observers. The direct technical effects are: first, it makes dynamic estimation of tire radial stiffness, which cannot be directly measured, possible; second, the high-precision model significantly reduces estimation errors caused by model mismatch, improving the overall system's sensitivity to tire pressure changes and estimation accuracy; finally, this model, combined with an extended Kalman filter incorporating an adaptive forgetting mechanism, achieves rapid, stable, and accurate estimation and identification of gradual tire pressure changes and sudden tire blowouts, fundamentally improving the performance of indirect tire pressure monitoring systems.

[0070] S2. Based on the parameters of the dynamic model, a state observer is designed, and two forgetting factors are introduced to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer.

[0071] Specifically, addressing the challenge of balancing rapid tracking and steady-state accuracy in traditional state observers during highly time-varying and fast-dynamic processes like tire pressure surges (such as tire blowouts), a novel adaptive extended Kalman filter (AFEKF) observer with dual forgetting factors and collaborative optimization is designed. Its core principle is as follows: First, the forgetting factor dynamically adjusts the prior estimation error covariance, actively reducing the algorithm's dependence on outdated observation data, thereby enhancing the observer's sensitivity to state surges. Simultaneously, the second forgetting factor drives online adaptive estimation of measurement noise statistics, enabling the observer to correct its perception of current measurement uncertainties in real time, suppressing noise interference and improving steady-state estimation accuracy. These two factors work together to achieve an adaptive optimal balance between the observer's dynamic and steady-state performance under different operating conditions.

[0072] Specifically, an Extended Kalman Filter (EKF) framework is designed based on the aforementioned high-precision nonlinear state-space model. First, the Jacobian matrix of the state transition equation is calculated in each sampling period to locally linearize the system. The key optimization step lies in a dual improvement to the standard EKF recursive formula: First, in the prediction step, a constant forgetting factor greater than 1 is introduced into the recursive formula for the prior estimation error covariance matrix. This allows the covariance matrix to be appropriately amplified, thereby increasing the Kalman gain and assigning higher weights to the innovation, thus accelerating the observer's response to state changes. Secondly, in the update step, a second forgetting factor λ2 (0 < λ) is introduced. <1) and weighting coefficients d k An improved Sage-Husa adaptive estimation algorithm is constructed. Based on real-time innovation sequences, this algorithm recursively estimates and updates the mean and covariance matrix of measurement noise online, enabling the observer to automatically adapt to measurement uncertainty fluctuations caused by changes in sensor characteristics or environmental interference, thereby effectively suppressing steady-state errors.

[0073] Furthermore, two forgetting factors and It is the core adjustable parameter. The value is slightly greater than 1 (e.g., in the range of 1.005 to 1.05), and its magnitude directly affects the trade-off between the response speed of state tracking and the stability of filtering. The value, ranging from 0 to 1 (e.g., 0.95 to 0.99), determines the "memory length" and adaptive speed of the measurement noise statistics estimation. The closer the value is to 1, the greater the weight of historical data, resulting in a smoother estimate but slower adaptability. Weighting coefficients d kThe design enables the algorithm to learn quickly in the initial stage and then converge asymptotically, ensuring the convergence and robustness of noise estimation. These parameters together constitute a fine-tuning mechanism for the observer's performance.

[0074] Specifically, this adaptive observer is designed to handle complex dynamics and disturbances during vehicle operation. It is suitable not only for routine monitoring of slow tire pressure changes but also for effectively handling sudden, abrupt changes in tire pressure caused by a tire blowout. When the vehicle experiences acceleration, braking, or steering that alters the dynamic characteristics of the suspension, its adaptive noise estimation function compensates for model errors and disturbances, ensuring the continuity of tire pressure estimation. This design enables the system to operate stably in various real-world driving and tire blowout scenarios, including high-speed straight driving, variable-speed steering, and different levels of road roughness.

[0075] Specifically, simulations have verified that the AFEKF observer proposed in this invention exhibits significant advantages over conventional EKF and FEKF with only a single forgetting factor. Under single-wheel blowout conditions, the tire pressure estimate converges to the true value within 0.35 seconds, with an average reduction of 63.1% in root mean square error (RMSE) and 59.9% in peak-to-peak fluctuation after convergence. In robustness tests involving multiple blowouts, shifting steering, and different road surface conditions, the observer maintains high accuracy and low fluctuation. Its estimation accuracy advantage is particularly prominent under adverse road conditions, effectively resolving the contradiction between dynamic tracking and steady-state accuracy in traditional methods. This provides a core algorithm guarantee for highly reliable real-time tire pressure estimation and blowout identification.

[0076] Furthermore, S2 includes:

[0077] S21, using the forgetting factor Reduce the observer's reliance on historical observation data:

[0078] ;

[0079] in, A constant forgetting factor. Here is the state transition matrix. The process noise covariance matrix is... The prior state estimation error covariance matrix is... The posterior state estimation error covariance matrix is... for The transpose of .

[0080] Specifically, using the forgetting factor The technique of amplifying the prior estimation error covariance matrix aims to overcome the inherent limitation of the standard Extended Kalman Filter (EKF) algorithm, which suffers from delayed response to sudden changes in system state due to its over-reliance on historical data. This is achieved by introducing a scaling factor greater than 1 into the recursive formula. This artificially increases the uncertainty of the prior estimate at the current moment (i.e., According to the Kalman gain calculation formula, this directly results in the gain matrix... The increase in Kalman gain means that the state update step will impart new information about the measurement at the current moment. Higher weights enable the observer to correct its state estimates more quickly based on the latest sensor information, significantly improving its ability to track rapidly changing dynamics such as sudden changes in tire pressure. The essence of this principle is to actively manage the "memory" and "confidence" of the filter, strategically reducing reliance on past estimates when a sudden change in state is possible, and instead responding more sensitively to newly emerging observational evidence.

[0081] Furthermore, this step is integrated into the prediction (time update) stage of the Adaptive Extended Kalman Filter (AFEKF). In each sampling period... The algorithm first uses the posterior estimate from the previous time step. and system input Through the state transition function The prior state estimate is calculated. Next, the Jacobian matrix of the state transition equation is calculated. This achieves local linearization. The key improvement lies in the calculation of the prior estimation error covariance matrix. In this case, the standard formula is not used directly. Instead, Multiply the term by a constant forgetting factor (in ), plus the process noise covariance matrix Here This is the updated posterior estimation error covariance matrix from the previous time step. After this operation, the magnified [matrix / covariance matrix ... It was used in subsequent Kalman gain and state update calculations to achieve targeted enhancement of the filter's dynamic response.

[0082] Furthermore, the forgetting factor This is the core adjustable parameter of the optimization measure, and its value directly affects the balance between the observer's dynamic performance and stability. Typically, It is set to a constant value slightly greater than 1, for example, selected within the range of 1.005 to 1.05. If While excessively large values ​​can significantly improve tracking speed, they can also over-amplify estimation uncertainty, potentially leading to excessively high filter gain, introducing excessive measurement noise, and causing drastic fluctuations or even instability in the estimation results; conversely, if... If the value is too close to 1, the optimization effect is not obvious and it is difficult to cope with rapid changes in tire pressure. Therefore, The determination of the filter needs to be carried out through simulation and experiment. Under the premise of ensuring stable convergence of the filter, the fine tuning should be carried out with the goal of achieving rapid tracking of the pressure change rate of a typical tire blowout (e.g., a loss of more than 25% of pressure within 1 second).

[0083] Specifically, this invention optimizes the most challenging scenario in indirect tire pressure monitoring: instantaneous tire blowout detection. When a tire blowout occurs while the vehicle is traveling at high speed, the tire pressure drops rapidly within tens to hundreds of milliseconds, representing a typical step-like state change. Traditional state observers are slow to track such changes, resulting in delayed warnings. This invention introduces… The AFEKF observer, through the aforementioned mechanism, can immediately increase the sensitivity to feedback from the suspension acceleration sensor after such a sudden event occurs, driving the tire pressure estimate to quickly deviate from its original steady state and rapidly track and converge to the true pressure value after the abrupt change. This buys valuable time for the subsequent tire blowout detection algorithm and is one of the key algorithmic components for achieving the "rapid" early warning function.

[0084] Specifically, simulation tests of the embodiments have verified that this improvement is superior to the one without the introduction of [the previous method]. The standard EKF (Effective Knock Factor) can reduce the convergence time of tire pressure estimation from untrackable or slow tracking to less than 1 second (e.g., 0.35 seconds) in the event of a single-wheel blowout, achieving rapid capture of pressure mutations. Simultaneously, through the integration with a second forgetting factor... Combined with adaptive noise correction, the system can quickly recover a high-precision steady-state estimate after rapid tracking, avoiding the problem of sacrificing accuracy for speed. This technical solution effectively resolves the contradiction between "rapid change" and "accuracy" in tire pressure monitoring, improving the overall performance and reliability of the system under extreme conditions.

[0085] S22, by introducing an improved Sage-Husa adaptive filter, dynamically corrects the measurement noise covariance matrix. :

[0086] ；

[0087] in, These are weighting coefficients. A constant forgetting factor. , For the innovation vector, , Estimate the measurement noise covariance matrix and mean at the current moment. For prior state estimation, The original observation residuals, For the estimation of the measurement noise covariance matrix at the next time step, Let be the current information vector. For the new information vector transpose, To estimate the mean vector of the measurement noise at the next time step, This represents the actual sensor measurement at the current moment. For the observation model function, For the estimation of the prior state at the current moment, For the system input at the current moment, Forgetting factor The kth power.

[0088] Specifically, considering the potential for unknown or time-varying statistical characteristics of measurement noise during vehicle operation (such as sensor performance fluctuations and road impact interference), an improved Sage-Husa adaptive filter is introduced to overcome the estimation bias or divergence problems caused by the fixed noise covariance matrix in traditional Extended Kalman Filter (EKF). Its core principle is to utilize the real-time "information" sequence generated during the filtering process—the difference between the observed actual value and the model's predicted value—as an effective information source reflecting the current measurement uncertainty. By designing a recursive algorithm with fading memory characteristics, the mean of the measurement noise is estimated and updated online and dynamically. With covariance matrix This allows the state observer to continuously "learn" and adapt to the actual measurement environment. This principle ensures that the filter's perception of noise statistics remains synchronized with the real situation, thereby effectively suppressing steady-state errors introduced by model mismatch or environmental interference, and improving estimation accuracy and robustness.

[0089] Furthermore, this adaptive correction mechanism is embedded in the update phase of the AFEKF. After calculating the Kalman gain... With the new vector Then, perform the following recursive update: First, calculate the time-varying weighting coefficients. Secondly, utilize Update the covariance matrix estimate of the measurement noise: This formula shows that the new estimate It is the estimate from the previous moment. Autocorrelation matrix with current information A convex combination, the weight of which is determined by Dynamic adjustment. Simultaneously, the mean of the measured noise is updated similarly: This is to correct for possible non-zero mean bias. Updated and The filtering calculations used in the next sampling period will form a closed-loop adaptive adjustment.

[0090] Furthermore, the second forgetting factor It is a key parameter for controlling adaptive speed and smoothness, and its value range is limited to... . The closer the value is to 1 (e.g., 0.99), the higher the weighting coefficient. Follow The larger the noise level and the slower the decay, the longer the algorithm retains historical data, resulting in smoother noise estimation but slower adaptation; conversely, If the value is small (e.g., 0.95), then The decay is faster, the algorithm is more sensitive to recent information, and the adaptive speed is faster, but the estimation fluctuation may increase. Weighting coefficients The design ensures that the algorithm is effective in the initial stage ( When the noise level is low, it has a large learning rate, which can quickly approximate the true noise statistics; as time goes on ( (increase) As the value gradually decreases, the updates become smoother, ensuring the asymptotic convergence and stability of the estimation.

[0091] Specifically, in actual operation, the road surface grades experienced by vehicles (such as from smooth highways to rough asphalt roads), driving operations (such as sudden changes in vehicle posture due to emergency braking and steering), and sensor operating conditions may all change, resulting in the statistical characteristics of measurement noise not being constant. This method enables the observer to automatically adapt to these changes, whether on ISO-A standard roads, more challenging B-grade roads, or actual bumpy roads, through online correction. This allows for accurate characterization of current measurement uncertainties. This enables the tire pressure estimation system to maintain superior filtering performance and estimation accuracy even when faced with non-stationary disturbances.

[0092] Specifically, compared with the single forgetting factor FEKF, AFEKF, due to its adaptive noise correction capability, significantly reduces the root mean square error (RMSE) and peak-to-peak value (PP) fluctuations of its tire pressure estimation results under various test conditions. Particularly in robustness tests with changing road conditions, when the road surface changes from Class A to Class B or is a measured bumpy road surface, the estimation error of the FEKF algorithm amplifies sharply, while the AFEKF algorithm proposed in this invention still maintains high accuracy and low fluctuations, with an RMSE optimization margin of up to 83%. This fully demonstrates the crucial role of dynamically correcting the measurement noise covariance matrix in suppressing environmental interference and improving the long-term reliability of the system under non-ideal conditions, and is one of the core technical guarantees for achieving highly robust tire pressure monitoring.

[0093] S3 monitors tire pressure data in real time and estimates tire pressure results through a state observer. Based on the tire pressure results and the state observer, it generates a dimensionally unified information squared burst identification algorithm to achieve rapid detection and early warning of bursts.

[0094] Specifically, when the vehicle is driving normally and the tire pressure matches the dynamic model, the "inspiration" sequence (i.e., the difference between the actual sensor measurements and the model predictions) generated by the state observer should follow a Gaussian distribution with a mean of zero and a theoretical inspiration covariance matrix as its statistical characteristic. Once a tire blowout occurs, the tire radial stiffness and tire pressure change abruptly, causing the system's true dynamic characteristics to deviate instantaneously from the model upon which the observer is based. This results in a significant change in the statistical characteristics of the inspiration sequence, with its amplitude increasing abnormally. By constructing and monitoring a scalar index reflecting the statistical characteristics of the inspiration—the dimensionlessly normalized inspiration square (NIS)—in real time, and comparing it with a pre-set statistical threshold, it is possible to effectively detect whether abrupt changes have occurred in the system model, thereby determining whether a tire blowout has occurred.

[0095] Specifically, the algorithm is executed synchronously after each update step of the Adaptive Extended Kalman Filter (AFEKF). First, it utilizes the innovation vector already computed at the current time step. and its theoretical covariance matrix Calculate the dimensionless square of the new information This indicator Statistically, it follows a chi-square distribution with degrees of freedom equal to the measurement dimension. Then, this calculated value... With a preset fixed threshold Real-time comparison is performed. Based on extensive simulations and prior analysis of typical tire blowout conditions, this embodiment sets the threshold to 200. The threshold is set if and only if the real-time calculation... When the value continuously exceeds the threshold, the algorithm determines that a tire pressure loss event has occurred that is sufficient to cause a sudden change in the vehicle's dynamic characteristics, i.e., a tire blowout. It then triggers a high-level warning signal, providing critical sensing input to the vehicle stability control system.

[0096] Furthermore, core decision threshold The choice of threshold is crucial in balancing the false alarm rate and the false negative rate. The threshold is set at 200, based on offline statistical analysis of a large amount of simulation data including normal driving, slight pressure fluctuations, and tire blowouts of varying degrees. This threshold corresponds to the value under normal model matching conditions. The probability of the indicator exceeding this value is extremely low (i.e., the significance level is extremely low), thus ensuring an extremely low false alarm rate. Meanwhile, under typical tire blowout conditions involving pressure loss (e.g., over 25%), The value will quickly and significantly exceed this threshold, ensuring an extremely high detection rate and rapid response. This fixed threshold setting provides the algorithm with clear and repeatable engineering judgment criteria, eliminating the need for complex online learning and guaranteeing the real-time nature and reliability of decision-making.

[0097] Specifically, this tire blowout recognition algorithm, as the final decision-making output module of the entire tire pressure monitoring system, directly serves the vehicle's active safety system. Its core application scenarios focus on conditions where the consequences of a tire blowout are severe, such as high-speed driving. Whether it's a sudden blowout in a single tire or multiple tires experiencing pressure loss sequentially or simultaneously, the algorithm can effectively identify the issue based on unified statistical criteria. The algorithm does not rely on the historical trend of slow changes in tire pressure estimates, but rather keenly captures the statistics provided by the AFEKF observer, reflecting the instantaneous degree of system mismatch. Therefore, it is particularly suitable for detecting instantaneous step changes in tire pressure, realizing a functional leap from "pressure estimation" to "event recognition", and providing a direct and rapid triggering condition for subsequent audible and visual alarms, instrument prompts or emergency intervention of the chassis control system.

[0098] Specifically, under the condition of a tire blowout with a single-wheel pressure loss of 25%, The value immediately spikes after a sudden change in tire pressure, far exceeding the threshold of 200, achieving instantaneous detection of tire blowouts with near-zero warning delay. Compared to simple methods that rely solely on whether the estimated tire pressure falls below a certain absolute threshold, this method utilizes rich multi-channel information and statistical data, exhibiting stronger anti-interference capabilities and effectively distinguishing between transient interference caused by severe road impacts and genuine tire blowout events. This algorithm, together with the aforementioned high-precision modeling and adaptive observer, constitutes a complete "perception-estimation-decision" chain. Ultimately, this enables the entire system to achieve continuous high-precision tire pressure estimation and ensure rapid and reliable early warning of tire blowouts, significantly improving the technical level of vehicle driving safety barriers, even in complex dynamic environments.

[0099] Furthermore, S3 includes:

[0100] S31, calculate the dimensionless innovation square.

[0101] Specifically, the core of this step lies in utilizing the innovation vector. Covariance matrix of the new information theory To construct a dimensionless statistic based on the relationship between them. This enables standardized evaluation of measurement residuals.

[0102] Furthermore, new information Defined as actual measured value Compared with prior state estimation Predicted output value The difference between them, i.e. ,in To measure the noise mean. The covariance matrix of the innovation theory. The measurement equation of the state observer, Jacobian matrix With the prior error covariance matrix and the measurement noise covariance matrix Together they form the expression: This matrix reflects the theoretical distribution characteristics of the innovation at the current moment, based on the system model and noise statistics.

[0103] Furthermore, the dimensionless square of the new information The calculation formula is Its essence is to place the innovation vector in The statistic is normalized in the covariance space to eliminate dimensional differences between different measurement channels. This statistic follows a chi-square distribution. Its degrees of freedom are equal to the dimension of the measurement vector. In practical applications, a threshold for the chi-square distribution is set. ,when At that time, it was determined that a tire blowout had occurred.

[0104] Furthermore, The calculation depends on and Real-time updates, and An improved Sage-Husa adaptive algorithm is then used for online adjustment to address the time-varying characteristics of measurement noise. In this embodiment, the following settings are configured: The threshold value for tire blowout detection is calibrated based on the statistical characteristics of simulation data and the robustness requirements of the actual system.

[0105] Specifically, this step plays a crucial role in the vehicle tire pressure monitoring system. By utilizing the statistical properties of standardized information vectors, it improves the sensitivity to sudden changes in tire pressure while reducing the false alarm rate. Even under complex conditions such as multiple tire blowouts and sudden steering changes, this method maintains high recognition accuracy and stability, providing timely and reliable blowout warnings to the self-stabilizing controller, thereby enhancing vehicle driving safety.

[0106] S32, Set the tire blowout detection threshold to 200, when the calculated value... When the threshold is exceeded, a tire blowout is determined to have occurred and a warning signal is triggered.

[0107] Specifically, this step, based on the state estimation error information output by the Adaptive Forgetting Extended Kalman Filter (AFEKF) algorithm, uses the dimensionally normalized innovation square (NIS) to achieve rapid identification of tire blowout events, which is a key link in the entire system to realize fault detection and early warning.

[0108] Furthermore, the dimensionless square of the new information The calculation formula is ,in Let be the innovation vector, representing the difference between the actual measured value and the predicted value. Let be the theoretical covariance matrix of the innovation. This formula originates from the statistical characteristic analysis of state estimation error in the AFEKF algorithm, and introduces an adaptive mechanism. It can dynamically adjust, thereby improving sensitivity to abnormal states. When Exceeding the set threshold When the value is measured, it indicates that there is a significant deviation between the current measured value and the model prediction value, which may be caused by a sudden change in tire stiffness due to a tire blowout. Based on this, the system determines that a tire blowout has occurred and triggers a warning signal.

[0109] Furthermore, the tire blowout detection threshold Set to 200, this value was determined based on statistical analysis of extensive simulation data and actual test results to ensure that it does not trigger false triggers within the normal tire pressure fluctuation range, while being able to quickly identify a blowout. (New Information Covariance Matrix) The calculation depends on the Jacobian matrix at the current time step. Prior error covariance matrix and the measurement noise covariance matrix These matrices are updated recursively in the AFEKF algorithm, exhibiting real-time performance and adaptability. Furthermore, the new information... The calculation needs to take into account the mean of the measurement noise. To improve the robustness of recognition.

[0110] Specifically, this step is applicable to high-speed vehicle operation, especially under conditions of significant road disturbance and sudden tire changes. By real-time acquisition of acceleration signals of sprung and unsprung masses in the suspension system, combined with tire stiffness estimates output by the AFEKF algorithm, the system can quickly identify anomalies and issue warnings at the moment of a tire blowout (such as a 25% drop in tire pressure), providing timely input signals to the vehicle's self-stability control system and thus improving vehicle driving safety under tire blowout conditions.

[0111] Specifically, setting As a blowout detection threshold, this threshold can effectively distinguish between normal tire pressure fluctuations and sudden stiffness changes caused by a blowout. In simulation tests, this threshold demonstrated excellent recognition performance under complex conditions such as ISO-A level random road surfaces, multiple tire blowouts, and variable steering, significantly improving the response speed and accuracy of blowout detection. This step not only enhances the real-time performance of the indirect tire pressure monitoring system but also provides a reliable basis for the execution of subsequent control strategies, possessing significant engineering application value.

[0112] S4 is a dynamic compensation step based on the vehicle's shifting and steering conditions.

[0113] Specifically, this step compensates for the dynamic behavior of the vehicle during speed change and steering by introducing an Adaptive Forgetting Extended Kalman Filter (AFEKF), thereby effectively addressing the model uncertainty in tire stiffness estimation caused by changes in the vehicle's motion state.

[0114] Furthermore, AFEKF introduces a forgetting factor on top of the Extended Kalman Filter (EKF). An adaptive noise covariance update mechanism is used to dynamically weight historical state information. In the state prediction phase, prior state estimation... From the state transition equation The calculation shows that, among which This represents the input variables at the current moment, such as road disturbances and steering wheel angle. Then, the Jacobian matrix is ​​used... Linearize the state transition function, where The sampling time interval is 1 ms. This is the identity matrix. In the update of the state estimation error covariance, a forgetting factor is used. The covariance matrix is ​​updated with a weighted average, and the formula is as follows: ,in The process noise covariance matrix represents the modeling error.

[0115] Furthermore, during the measurement update phase, AFEKF uses the Jacobian matrix... Linearize the measurement function and calculate the Kalman gain. ,in To measure the noise covariance matrix. (News) Used to assess the deviation between the current estimate and the actual measurement, where To measure the noise mean, further updates are performed using an adaptive mechanism. Using weighted coefficients （ ） The new information covariance is dynamically adjusted using the following formula: This improves the steady-state estimation accuracy of the filter under complex operating conditions such as speed change and steering.

[0116] Specifically, in this invention It is usually set to 1.1~1.5 to control the degree of forgetting of historical data; The value is set to 0.8~0.95 to adjust the update rate of the measurement noise covariance. The sampling frequency is 1 kHz to meet the real-time requirements of vehicle dynamic response.

[0117] Specifically, this step is applicable to tire pressure estimation and blowout identification under complex driving conditions such as gear shifting, steering, and braking. Through dynamic compensation, AFEKF can effectively suppress estimation bias caused by changes in vehicle motion state, improving the robustness and real-time performance of tire pressure estimation.

[0118] Specifically, this step significantly improves the convergence speed and steady-state accuracy of tire pressure estimation under variable speed steering conditions. For example, under ISO-A grade road conditions, a vehicle speed of 80 km / h, and a single-wheel tire blowout condition, AFEKF achieves a RMSE optimization of 63.1% and a peak-to-peak optimization of 59.9%, effectively enhancing the reliability and response speed of blowout detection.

[0119] Furthermore, S4 includes:

[0120] S41, during vehicle braking, deceleration, and steering, adjusts the input variables of the state transition equation based on real-time vehicle speed and steering angle.

[0121] Specifically, the core of this step lies in dynamically adjusting the input variables. This is to reflect the actual motion state of the vehicle under complex working conditions, thereby improving the modeling accuracy of the state transition equation for the vehicle's dynamic behavior.

[0122] Furthermore, input variables Includes vehicle lateral acceleration and longitudinal acceleration These two variables are important factors affecting the estimation of tire radial stiffness in the suspension system dynamics model. During braking or steering, the vehicle's acceleration distribution changes significantly, thus affecting the contact force characteristics between the tire and the ground. Therefore, this invention uses an onboard IMU (Inertial Measurement Unit) to collect data in real time. and Data, combined with vehicle speed and steering angle Dynamic compensation is performed. Specifically, when the vehicle is turning, the lateral acceleration... With steering angle The relationship is nonlinear, and its calculation can be based on the vehicle's kinematics model, such as... ,in For the turning curvature, by It is derived from parameters such as wheelbase. Longitudinal acceleration The accelerometer measures the noise directly and filters it during braking to eliminate transient impact noise.

[0123] Furthermore, vehicle speed The sampling frequency is typically set to 100 Hz, and the steering angle is... The sampling frequency is 50 Hz to ensure the real-time performance and stability of the input variables. Lateral acceleration. and longitudinal acceleration The measurement error should be controlled within 0.1. The input values ​​must be within the range of g to meet the state estimator's requirements for input accuracy. Furthermore, to improve model robustness, the input variables need to undergo low-pass filtering with a cutoff frequency set to 5 Hz to suppress high-frequency noise interference.

[0124] Specifically, this step applies to tire pressure estimation and blowout detection when a vehicle is braking, decelerating, or steering in complex road conditions. For example, in situations such as high-speed lane changes, emergency braking, or cornering, the vehicle's acceleration distribution changes drastically; dynamic adjustments are made to... This can effectively improve the sensitivity of the AFEKF algorithm to changes in tire radial stiffness, thereby accelerating the convergence speed of tire pressure estimation and improving the accuracy of tire blowout identification.

[0125] Specifically, this step enhances the adaptability of the state transition equation to vehicle dynamics by incorporating real-time vehicle speed and steering angle information, thereby improving the estimation accuracy and response speed of the AFEKF algorithm under non-steady-state conditions. Experiments show that under variable-speed steering conditions, AFEKF's tire pressure estimation RMSE is optimized by an average of 42.0% compared to FEKF, and the peak-to-peak (PP) error is optimized by 46.3%, significantly improving the system's practicality and reliability in complex driving scenarios.

[0126] S42 converts the dynamically estimated tire radial stiffness into a tire pressure estimate to improve the accuracy of tire pressure estimation under variable speed steering conditions.

[0127] Specifically, the principle of this invention is based on the strong correlation between inflation pressure and radial stiffness in tire mechanics. Under a given tire structure, its radial stiffness is primarily affected by the internal inflation pressure. Within a certain pressure range, the tire's radial stiffness and inflation pressure exhibit an approximately monotonically linear relationship. Therefore, by dynamically estimating the tire's real-time radial stiffness with high precision and frequency, the corresponding absolute tire pressure value can be derived from this intrinsic relationship. This principle overcomes the limitation of traditional indirect tire pressure monitoring systems, which can only identify relative tire pressure changes, providing a theoretical basis for the quantitative and rapid estimation of the absolute tire pressure of all four wheels of a vehicle.

[0128] Specifically, firstly, in the offline phase, the static radial stiffness of the target tire was measured under different standard inflation pressures (e.g., 186.2 kPa to 296.5 kPa) using a tire testing bench, obtaining a series of pressure-stiffness sample pairs. The dataset was then fitted using linear regression with the least squares method to obtain the calibration transformation formula: ,in This is an estimated tire pressure value. The tire radial stiffness is estimated in real time by the AFEKF algorithm. and These are calibration coefficients determined through experimental data. During online operation, the AFEKF algorithm outputs estimates of the radial stiffness of the four wheels in each sampling period. The estimated value is then substituted into the calibration formula in real time to calculate the corresponding absolute tire pressure estimate. This process runs synchronously with the state estimation algorithm, seamlessly integrated to ensure the real-time nature of tire pressure estimation.

[0129] Furthermore, this conversion method aims to directly transfer the accuracy advantage of radial stiffness estimation to tire pressure estimation. Based on simulation verification, under variable-speed steering conditions including braking and steering, the root mean square error (RMSE) of tire pressure estimation output by the AFEKF algorithm is optimized by an average of approximately 42.0% compared to the traditional FEKF method, and the peak-to-peak fluctuation is optimized by an average of approximately 46.3%. Specifically, under the condition of braking and steering at an initial speed of 80 km / h, the RMSE of tire pressure estimation for the blown tire can be as low as 0.0840 bar, and the pressure estimation curves of all tires are smooth, which can quickly and stably track the sudden change of the true tire pressure value without significant overshoot or lag, meeting the stringent requirements for estimation accuracy and response speed under high-speed dynamic conditions.

[0130] Furthermore, this estimated tire pressure is directly input into the vehicle's advanced chassis control system as a key vehicle status parameter. Especially after the blowout detection module determines a blowout has occurred, the accurate and rapid estimate of tire pressure loss and its gradient can serve as crucial feedforward information, providing real-time updates to systems such as Electronic Stability Control (ESC), Adaptive Cruise Control (ACC), and emergency stability control strategies after a blowout. Application scenarios cover highway driving at constant speeds, urban road start-stop and shifting, cornering, and the resulting blowout emergency situations, providing continuous tire pressure status awareness for driving safety across all scenarios.

[0131] Specifically, the stiffness-tire pressure conversion method has achieved significant technical effects. First, it enables quantitative estimation of absolute tire pressure, overcoming the fundamental shortcomings of indirect tire pressure monitoring systems. Second, it significantly improves the estimation accuracy and robustness under complex dynamic conditions, especially under variable speed steering, solving the industry-wide problem of estimation inaccuracies caused by drastic changes in vehicle motion. Third, the conversion process is based on clear physical relationships and experimental calibration, making the algorithm reliable, computationally efficient, and easy to embed into vehicle electronic control units. Fourth, it provides high-confidence real-time tire pressure information input for tire blowout detection and subsequent autonomous vehicle stability control, forming a "perception-decision-control" safety closed loop, fundamentally enhancing the vehicle's active safety performance.

[0132] This invention discloses an adaptive forgetting extended Kalman filter tire pressure monitoring method. By constructing a precise dynamic framework integrating a high-fidelity suspension model and genetic algorithm parameter identification, and designing an adaptive state observer with dual forgetting factors for collaborative optimization, it effectively solves the problems of low estimation accuracy, large tracking delay, and poor robustness caused by model mismatch and noise interference in traditional indirect tire pressure monitoring methods under complex dynamic and abrupt conditions. It achieves integrated intelligent monitoring from precise mechanism modeling and adaptive filtering estimation to rapid tire blowout identification based on statistical decision-making, significantly improving the accuracy of tire pressure estimation, real-time warning, and overall system reliability in complex scenarios such as single / multi-wheel blowouts, variable steering, and different road surface grades.

[0133] Example 2

[0134] To achieve the above invention, embodiments of the present invention also provide another adaptive forgetting extended Kalman filter tire pressure monitoring method, including:

[0135] This method's technical approach includes suspension system dynamics model and parameter identification, a fast indirect tire pressure estimation algorithm for the whole vehicle based on adaptive forgetting extended Kalman filtering, and a tire blowout identification algorithm based on dimensionally unified innovation squares, etc. The technical approach is as follows: Figure 2 As shown.

[0136] In one embodiment of the present invention, a dynamic model and parameter identification of a suspension system are provided.

[0137] Specifically, high-precision mechanistic models can significantly improve the accuracy of vehicle state and parameter estimation. First, a seven-DOF model of the vehicle suspension is built; second, the stiffness of the four wheels of the vehicle is extended into state variables, and nonlinear state-space equations are derived; finally, a genetic algorithm is used to identify the inherent parameters of the suspension model, making the model more accurately reflect the nonlinear dynamic characteristics of the actual vehicle and providing more accurate dynamic model parameters for the subsequent design of the state observer.

[0138] Furthermore, such as Figure 3 The figure shows the dynamic model of the vehicle's seven-degree-of-freedom suspension system. ， 、 、 、 These represent the unsprung masses at the front left, front right, rear left, and rear right, respectively. 、 、 、 These represent the road surface unevenness inputs for the left front, right front, left rear, and right rear, respectively. 、 、 、 These represent the vertical displacements of the unsprung masses at the left front, right front, left rear, and right rear, respectively. 、 、 、 These represent the vertical displacements of the sprung masses at the left front, right front, left rear, and right rear, respectively. 、 、 、 These represent the damping coefficients of the suspension shock absorbers at the front left, front right, rear left, and rear right, respectively. 、 、 、 These represent the suspension spring stiffness for the front left, front right, rear left, and rear right, respectively. 、 、 、 These represent the radial stiffness of the tires at the front left, front right, rear left, and rear right, respectively. Indicates the sprung mass. ， These represent the pitch and roll moments of inertia of the vehicle body, respectively. This indicates the vertical displacement of the vehicle's center of gravity. Indicates the pitch angle displacement of the vehicle body. Indicates the body roll angle displacement. 、 These represent the distances from the vehicle's center of gravity to the front and rear axles, respectively. ， 、 These represent the distances from the vehicle's center of gravity to the left and right wheels, respectively.

[0139] Specifically, based on the Newton-Euler equations, without considering the variable speed condition, and only considering the vertical, pitch, roll motions of the vehicle body and the four-wheel bounce caused by the unevenness of the road surface, the differential equations of the suspension system dynamics are constructed as follows:

[0140]

[0141] In the formula, 、 、 、 The tables separately represent the suspension forces at the left front, right front, left rear, and right rear, and their calculation formulas are as follows:

[0142]

[0143] Assuming the vehicle's pitch and roll angles are small during normal driving, then:

[0144]

[0145] Equation (3) can be used to approximate the vertical displacements at the connection points between the vehicle's left front, right front, left rear, and right rear sides and the suspension as follows:

[0146]

[0147] In the formula, This represents the vehicle roll angle, in radians (rad). The selected state and output variables are as follows:

[0148]

[0149] From equation (5), and considering modeling error and measurement noise, equation (1) can be rewritten as the nonlinear state-space equation:

[0150]

[0151] Among them, the state transition equation Represented as:

[0152]

[0153] in, It has a mean of 0 and a covariance matrix of Process noise characterizes modeling error.

[0154] Measurement equation Represented as:

[0155]

[0156] in, It has a mean of 0 and a covariance matrix of Process noise characterizes measurement error.

[0157] Specifically, regarding such as Figure 3 The inherent parameters of the vehicle suspension model shown are identified. An ISO-A level random road surface is built in CarSim simulation software to simulate high-speed straight-line driving at a vehicle speed of 80 km / h and a sampling interval of 1 ms. Using the sprung and unsprung mass accelerations of the four wheels in CarSim as reference responses, minimizing the suspension model response error is used as the cost function. A genetic algorithm is employed for parameter optimization and identification. The designed cost function is as follows:

[0158]

[0159] In equation (9), The root mean square error of the acceleration response of the sprung mass at the left front, right front, left rear, and right rear is represented. The root mean square error of the unsprung mass acceleration response represents the left front, right front, left rear, and right rear, i.e.

[0160]

[0161] In equation (10), 、 These represent the sprung and unsprung mass acceleration responses measured in CarSim for the left front, right front, left rear, and right rear, respectively. 、 These represent the sprung and unsprung mass acceleration response results for the left front, right front, left rear, and right rear of the suspension model, respectively. The length of the sampled data.

[0162] In one embodiment of the present invention, a fast indirect tire pressure estimation algorithm for the whole vehicle is based on adaptive forgetting extended Kalman filtering.

[0163] Specifically, considering the rapid changes in tire pressure after a blowout, this invention employs a forgetting factor in the design of the state observer to weaken its dependence on historical observation data, thereby improving the observer's dynamic tracking capability. Furthermore, an improved Sage-Husa adaptive filter is introduced to dynamically correct the measurement noise covariance matrix. This optimizes the steady-state estimation accuracy of the state estimator. The specific steps of the AFEKF algorithm are as follows:

[0164] (1) Calculate the prior state estimate

[0165]

[0166] (2) Calculate the Jacobian matrix of the state transition equation.

[0167]

[0168] (3) Calculate the prior state estimation error covariance matrix

[0169]

[0170] in, A constant forgetting factor. .

[0171] (4) Calculate the Jacobian matrix of the measurement equation.

[0172]

[0173] (5) Calculate the Kalman filter gain

[0174]

[0175] (6) Calculate the new information

[0176]

[0177] Among them, new information Defined as the actual value of the measured variable Compared with the predicted value Measure the mean noise difference, To measure the mean noise level.

[0178] (7) Calculate the posterior state estimate

[0179]

[0180] (8) Update the posterior state estimation error covariance matrix

[0181]

[0182] (9) Update the measurement noise covariance matrix

[0183]

[0184] in, These are weighting coefficients. A constant forgetting factor. .

[0185] Specifically, the AFEKF algorithm was used to dynamically estimate the radial stiffness of the four wheels of the vehicle. Now, without considering the influence of environmental factors such as temperature and altitude on tire inflation pressure, the tire pressure estimation results are obtained by referring to the tire test bench data in Table 1 and the linear regression fitting formula (20).

[0186] Table 1. Test results of tire radial stiffness under different inflation pressures

[0187]

[0188]

[0189] in, Inflate the tires to the specified pressure.

[0190] In one embodiment of the present invention, a tire blowout recognition algorithm is based on dimensionally unified information squared.

[0191] Specifically, real-time monitoring of tire pressure and timely issuance of tire blowout warnings provide sufficient time margin for the intervention of the vehicle's self-stabilization controller after a tire blowout, optimizing control performance and mitigating the risk of secondary instability caused by delayed controller intervention. Considering the strong time-varying characteristics of vehicle state variables at the moment of a tire blowout, this invention proposes a tire blowout identification algorithm based on the dimensionlessly normalized innovation square (NIS) based on the overall tire pressure estimation results provided by the aforementioned AFEKF algorithm. A NIS threshold is set, and a tire blowout is considered to have occurred when the NIS exceeds the set threshold.

[0192] Furthermore, the formula for calculating the square of the dimensionless innovation is as follows:

[0193]

[0194] In equation (26), To achieve dimensionless information square, in this embodiment... , This is the covariance matrix of the innovation theory.

[0195] Another adaptive forgetting extended Kalman filter tire pressure monitoring method in this invention constructs a nonlinear state-space model that accurately reflects tire stiffness dynamics through a high-precision suspension dynamics model and genetic algorithm parameter identification. Furthermore, an adaptive estimation algorithm with a dual forgetting mechanism is used to achieve online correction of time-varying noise characteristics and rapid tracking of tire pressure fluctuations. A blowout identification mechanism based on innovation theory covariance significantly enhances sensitivity and robustness in detecting hazardous conditions. Ultimately, this method forms a complete technical solution from accurate modeling and online parameter estimation to real-time fault diagnosis, effectively solving the shortcomings of traditional indirect tire pressure monitoring in complex conditions (insufficient accuracy) and slow response to blowouts, providing highly reliable key state perception information for vehicle active safety systems.

[0196] Example 3

[0197] To achieve the above invention, embodiments of the present invention also provide an application scenario for the adaptive forgetting extended Kalman filter tire pressure monitoring method, including:

[0198] In this embodiment, a joint simulation test was conducted using the vehicle simulation software CarSim and the mathematical software MATLAB / Simulink. The test scenario was a high-speed straight-line tire blowout. Different blowout conditions were designed by changing the steering wheel angle, the number of blowout tires, the percentage of tire pressure loss, and the road surface grade to verify the effectiveness and feasibility of this study. The comparison methods are as follows:

[0199] Specifically, EKF: Conventional Extended Kalman Filter. Commonly used in estimator design for nonlinear systems, its measurement noise covariance matrix is ​​constant; FEKF: Extended Kalman Filter with Forgetting Factor. The forgetting factor is used to reduce dependence on historical observation data, and its measurement noise covariance matrix is ​​constant; AFEKF: The method proposed in this invention. It introduces a forgetting factor and an adaptive adjustment strategy for the measurement noise covariance matrix to optimize the estimator's dynamic tracking performance and steady-state estimation accuracy.

[0200] In one embodiment of the present invention, a single-wheel tire blowout occurs.

[0201] Specifically, an ISO-A grade random road surface was selected as the road unevenness input, and the vehicle traveled in a straight line at a constant speed of 80 km / h for 50 s. Based on the warning threshold of the Passenger Car Tire Pressure Monitoring System Test Method (GB26149), the front left wheel stiffness was set to a sudden change from the nominal value of 268000 N / m to 216232 N / m at the 20th s of the simulation, with a pressure loss of 25%. The simulation results of the single-wheel blowout condition are shown in Figures 4(a), 4(b), 4(c), and 4(d), and the tire pressure estimation results for the single-wheel blowout condition are shown in Table 2.

[0202] Table 2. Tire pressure estimation results under single-wheel blowout conditions

[0203]

[0204] In one embodiment of the present invention, a multi-tire blowout scenario is described.

[0205] Specifically, to discuss the estimation sensitivity of AFEKF under multi-tire blowout conditions, while keeping the road surface, vehicle speed, and simulation time constant, a four-wheel underinflation condition was set up. At 20 seconds into the simulation, the front left tire, front right tire, rear left tire, and rear right tire experienced pressure losses of 15%, 25%, 50%, and 90%, respectively. The simulation results for the multi-tire blowout condition are shown in Figures 5(a) and 5(b), and the steady-state estimation error results are shown in Table 3. Overall, even under multi-tire underinflation conditions, AFEKF still maintained a smaller signal fluctuation amplitude and higher steady-state estimation accuracy.

[0206] Table 3. Tire pressure estimation results under multiple tire blowout conditions

[0207]

[0208] In one embodiment of the present invention, the variable speed steering condition.

[0209] Specifically, considering only constant speed straight-line driving is insufficient to cover common driving conditions. Therefore, to simulate the aforementioned vehicle dynamic behavior, a scenario involving variable speed and steering, and a front left tire blowout, was set up. After braking and deceleration, the vehicle steered to the side of the road, maintaining a constant pressure loss. The simulation results for the variable speed and steering scenario are shown in Figures 6(a), 6(b), 6(c), and 6(d), and the tire pressure steady-state estimation error results are shown in Table 4. The results show that even under variable speed and steering driving conditions, AFEKF can accurately estimate tire pressure across the entire range.

[0210] Table 4. Tire pressure estimation results under variable steering conditions

[0211]

[0212] In one embodiment of the invention, robustness testing.

[0213] Specifically, when the road surface grade changes, the equivalent intrinsic parameters of the offline identified suspension model may change, thus affecting the estimation effect of the state estimator. The road surface was replaced with ISO-B grade random road surface and actual road surface, and the rest of the simulation settings were kept consistent with the single-wheel tire blowout condition to verify the robustness of the proposed algorithm. The simulation results are shown in Figure 7(a) and Figure 7(b), and the tire pressure steady-state estimation error results are shown in Table 5.

[0214] Table 5. Tire pressure estimation results for roads with varying road grades

[0215]

[0216] This invention presents an application scenario for an adaptive forgetting extended Kalman filter (ADKF) tire pressure monitoring method. This method effectively addresses the shortcomings of traditional indirect monitoring techniques, such as inaccurate estimation and delayed response to tire blowouts under complex dynamic conditions. It demonstrates excellent tire pressure estimation accuracy and dynamic tracking performance in various demanding scenarios, including high-speed straight driving, simultaneous underinflation of multiple tires, variable steering, and random road surfaces of varying grades. It also achieves rapid, sensitive, and reliable identification of tire blowout events. Its unique adaptive mechanism significantly improves the system's robustness to changes in road surface excitation, sensor noise, and model parameter uncertainties, providing timely and high-confidence critical state perception inputs for the active safety control system of vehicles experiencing tire blowouts, thereby comprehensively enhancing the vehicle's driving safety capabilities.

[0217] Example 4

[0218] To achieve the above invention, such as Figure 8 As shown, this embodiment also provides an adaptive forgetting extended Kalman filter tire pressure monitoring device 10, which includes:

[0219] The model building module 100 is used to build a seven-degree-of-freedom suspension dynamics model that includes the radial stiffness of the four tires. It selects a 14-dimensional state vector that includes the vertical displacement of the vehicle's center of gravity, pitch angle, roll angle, unsprung mass displacement of the four wheels, and their first derivatives. The radial stiffness of the four tires is augmented into system state variables to form an 18-dimensional state vector. An 8-dimensional output vector that includes the sprung mass acceleration and unsprung mass acceleration of the four wheels is defined. Based on the seven-degree-of-freedom suspension dynamics model, a nonlinear state-space equation is established. The inherent parameters of the seven-degree-of-freedom suspension dynamics model are identified using a genetic algorithm with CarSim simulation data as a reference, so as to construct more accurate dynamic model parameters.

[0220] The state observer design and optimization module 200 is used to design a state observer based on the parameters of the dynamic model, and to introduce two forgetting factors to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer.

[0221] The tire blowout condition identification module 300 is used to monitor tire pressure data in real time and estimate the tire pressure result through a state observer. Based on the tire pressure result and the state observer, a dimension-unified innovation squared tire blowout identification algorithm is generated to achieve rapid detection and early warning of tire blowout.

[0222] In one embodiment of the present invention, it further includes: a dynamic compensation module, used for dynamic compensation steps based on vehicle shifting and steering conditions: during vehicle braking deceleration and steering, adjusting the input variables of the state transition equation according to the real-time vehicle speed and steering angle; converting the dynamically estimated tire radial stiffness into a tire pressure estimate to improve the tire pressure estimation accuracy under shifting and steering conditions.

[0223] This invention discloses an adaptive forgetting extended Kalman filter tire pressure monitoring device. Through a model building module, it achieves high-precision suspension-tire coupled dynamics modeling and parameter identification. A dual forgetting factor collaborative mechanism is introduced through a state observer design and optimization module, achieving a balance between rapid tracking of sudden tire pressure changes and high-precision steady-state estimation. The tire blowout condition identification module performs statistical decision-making based on dimensionally unified information squares, enabling instantaneous detection and early warning of tire blowouts. Combined with a dynamic compensation module for adaptive correction of shift steering conditions, this device effectively solves the problems of estimation delay, decreased accuracy, and insufficient robustness caused by model mismatch, noise interference, and changes in operating conditions in traditional methods. It significantly improves the overall accuracy, real-time warning performance, and system reliability of tire pressure monitoring under complex driving environments.

[0224] To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 9 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads the executable program code stored in the memory 601 to run a program corresponding to the executable program code, so as to implement the various steps of the adaptive forgetting extended Kalman filter tire pressure monitoring method described above.

[0225] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements an adaptive forgetting extended Kalman filter tire pressure monitoring method as described in the foregoing embodiments.

[0226] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0227] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. An adaptive forgetting extended Kalman filter tire pressure monitoring method, characterized in that, include: S1. A seven-degree-of-freedom suspension dynamics model including the radial stiffness of the four tires is constructed. A 14-dimensional state vector is selected, which includes the vertical displacement of the vehicle's center of gravity, pitch angle, roll angle, unsprung mass displacement of the four wheels, and their first derivatives. The radial stiffness of the four tires is extended into system state variables to form an 18-dimensional state vector. An 8-dimensional output vector including the sprung mass acceleration and unsprung mass acceleration of the four wheels is defined. A nonlinear state-space equation is established based on the seven-degree-of-freedom suspension dynamics model. The inherent parameters of the seven-degree-of-freedom suspension dynamics model are identified using a genetic algorithm with CarSim simulation data as a reference, so as to construct more accurate dynamic model parameters. S2, Design a state observer based on the parameters of the dynamic model, and introduce two forgetting factors to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer; S3 acquires real-time sensor measurement data of sprung mass acceleration and unsprung mass acceleration corresponding to the 8-dimensional output vector. It dynamically estimates the radial stiffness of the four tires in the 18-dimensional state vector through the state observer and converts the estimated radial stiffness into a tire pressure estimate based on the pre-calibrated linear relationship between tire radial stiffness and tire pressure. It calculates the dimension-unified innovation square based on the innovation vector generated by the state observer. When the dimension-unified innovation square exceeds a preset threshold, it determines that a tire blowout has occurred and triggers an early warning.

2. The method as described in claim 1, characterized in that, Based on the aforementioned seven-degree-of-freedom suspension dynamics model, a nonlinear state-space equation is established, including: S11. Based on an 18-dimensional state vector and an 8-dimensional output vector, a nonlinear state-space model containing state transition equations and measurement equations is established. The state transition equations describe the relationship between the state vector and time, while the measurement equations describe the relationship between the output vector and the state vector and input vector.

3. The method as described in claim 1, characterized in that, Two forgetting factors are introduced to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer, including: S21, using the forgetting factor Reduce the observer's reliance on historical observation data: ； in, A constant forgetting factor. Here is the state transition matrix. The process noise covariance matrix is... The prior state estimation error covariance matrix is... The posterior state estimation error covariance matrix is... for Transpose of; S22, by introducing an improved Sage-Husa adaptive filter, dynamically corrects the measurement noise covariance matrix. : ； in, These are weighting coefficients. A constant forgetting factor. , For the innovation vector, , Estimate the measurement noise covariance matrix and mean at the current moment. For prior state estimation, The original observation residuals, For the estimation of the measurement noise covariance matrix at the next time step, Let be the current information vector. For the new information vector transpose, To estimate the mean vector of the measurement noise at the next time step, This represents the actual sensor measurement at the current moment. For the observation model function, Estimate the prior state at the current moment. For the system input at the current moment, Forgetting factor The k-th power.

4. The method as described in claim 1, characterized in that, A blowout detection algorithm based on tire pressure results and a state observer generates dimensionally unified information squares to achieve rapid blowout detection and early warning, including: S31, calculate the square of dimensionless innovation; S32, Set the tire blowout detection threshold to 200, when the calculated value... When the threshold is exceeded, a tire blowout is determined to have occurred and a warning signal is triggered.

5. The method as described in claim 1, characterized in that, Also includes: S4, Dynamic compensation steps based on vehicle shifting and steering conditions: S41, during vehicle braking, deceleration, and steering, adjust the input variables of the state transition equation according to the real-time vehicle speed and steering angle; S42 converts the dynamically estimated tire radial stiffness into a tire pressure estimate to improve the accuracy of tire pressure estimation under variable speed steering conditions.

6. An adaptive forgetting extended Kalman filter tire pressure monitoring device, characterized in that, include: The model building module is used to construct a seven-degree-of-freedom suspension dynamics model that includes the radial stiffness of the four tires. A 14-dimensional state vector is selected, which includes the vertical displacement of the vehicle's center of gravity, pitch angle, roll angle, unsprung mass displacement of the four wheels, and their first derivatives. The radial stiffness of the four tires is augmented into system state variables to form an 18-dimensional state vector. An 8-dimensional output vector is defined, which includes the sprung mass acceleration and unsprung mass acceleration of the four wheels. A nonlinear state-space equation is established based on the seven-degree-of-freedom suspension dynamics model. The inherent parameters of the seven-degree-of-freedom suspension dynamics model are identified using a genetic algorithm with CarSim simulation data as a reference, so as to construct more accurate dynamic model parameters. The state observer design and optimization module is used to design a state observer based on the parameters of the dynamic model, and introduce two forgetting factors to optimize the dynamic tracking capability and steady-state estimation accuracy of the state observer. The tire blowout condition identification module is used to acquire sensor measurement data of sprung mass acceleration and unsprung mass acceleration corresponding to the 8-dimensional output vector in real time. It dynamically estimates the radial stiffness of the four tires in the 18-dimensional state vector through the state observer, and converts the estimated radial stiffness into the tire pressure estimate based on the pre-calibrated linear relationship between tire radial stiffness and tire pressure. It calculates the dimension-unified innovation square based on the innovation vector generated by the state observer. When the dimension-unified innovation square exceeds the preset threshold, it determines that a tire blowout has occurred and triggers an early warning.

7. The apparatus as claimed in claim 6, characterized in that, Also includes: The dynamic compensation module is used for dynamic compensation steps based on vehicle shifting and steering conditions: during vehicle braking, deceleration, and steering, the input variables of the state transition equation are adjusted according to the real-time vehicle speed and steering angle; the dynamically estimated tire radial stiffness is converted into a tire pressure estimate to improve the tire pressure estimation accuracy under shifting and steering conditions.

8. An electronic device, characterized in that, include: processor; as well as A memory, coupled to the processor, is used to store computer programs; When the processor is configured to execute the computer program, it implements the adaptive forgetting extended Kalman filter tire pressure monitoring method as described in any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that, The system contains a computer program that, when executed by a processor, implements the adaptive forgetting extended Kalman filter tire pressure monitoring method as described in any one of claims 1 to 5.

Images