A vehicle-mounted suspension state estimation method and system fusing zero-mean virtual observation

Abstract

The application discloses a kind of fusion zero-mean virtual observation vehicle suspension state estimation method and system, it is related to vehicle suspension state estimation technical field, the sensor arranged on vehicle is used as actual measurement source;First, the whole vehicle suspension dynamics model considering road roughness excitation, vehicle lateral acceleration and longitudinal acceleration effect is established, and the observation equation is established by directly measurable vehicle body vertical acceleration, vehicle body pitch angular velocity and vehicle body roll angular velocity;Using the quasi-zero mean characteristic that long-term statistical expected mean of vehicle body displacement and suspension relative displacement approaches to zero, construct zero-mean virtual observation and introduce Kalman filter updating process, to estimate the state quantity required for subsequent suspension control in real time.The application can suppress integral drift under the condition of only relying on vehicle IMU actual measurement signal, improve the precision and stability of suspension state estimation, provide reliable state information for subsequent suspension control.

Description

Technical Field

[0001] This invention relates to the field of vehicle suspension state estimation technology, and in particular to a method and system for estimating vehicle suspension state by fusing zero-mean virtual observations. Background Technology

[0002] Suspension state estimation has always been an important research direction in the field of vehicle active control. Under the excitation of braking, steering, cornering, and road unevenness, the vertical, pitch, and roll of the vehicle body, as well as the relative motion at each suspension point, will change significantly, thus affecting vehicle ride comfort and stability. For semi-active suspension systems, state information such as vehicle attitude, suspension dynamic deflection, and their changing trends directly determine the distribution of damper control forces and the suspension response adjustment effect. Therefore, researching methods to obtain reliable suspension control state information under actual vehicle sensor configuration conditions is of great significance for improving the control performance and engineering application feasibility of semi-active suspensions.

[0003] Currently, existing technologies for acquiring suspension status typically employ either multi-sensor direct measurement methods or state estimation methods based on dynamic models. Multi-sensor direct measurement methods collect vehicle motion signals by configuring sensors for suspension displacement, suspension speed, and vehicle acceleration, providing the controller with relatively complete state input. Dynamic model-based state estimation methods, on the other hand, collect measurable vehicle motion signals and reconstruct the state variables required for control by combining the suspension dynamics model and observer algorithms. Compared to multi-sensor direct measurement methods, dynamic model-based state estimation methods are more suitable for suspension control applications with existing single-sensor onboard IMUs. However, these methods are susceptible to sensor noise, model parameter deviations, and vehicle motion coupling under complex operating conditions, making it difficult to stably provide reliable state information required for subsequent control.

[0004] To address the shortcomings of dynamic model-based suspension state estimation methods in noise suppression and state reconstruction, previous research has introduced Kalman filtering, augmented state estimation, or data-driven methods to improve the accuracy and robustness of suspension state estimation. These methods can improve the reconstruction effect of suspension control state information to some extent and can estimate control state variables that are difficult to measure directly using measurement signals. However, in real vehicles, when sensor configuration is limited, measurement noise exists, or the vehicle is in a braking-steering coupled condition, these methods are affected by integral accumulation error, model parameter deviation, and real-time application conditions, resulting in insufficient accuracy and stability of control state information.

[0005] The suspension system itself is characterized by nonlinearity, time-varying behavior, and coupling between vehicle motion and road excitation, making state estimation under actual vehicle operating conditions quite challenging. Inaccurate estimation of the state variables required for control will affect the subsequent judgments of the semi-active suspension controller regarding vehicle attitude and suspension response. Especially when relying solely on actual measurement signals from the onboard IMU, while the IMU can provide vehicle acceleration and angular velocity information, it cannot directly provide complete suspension control state information. How to utilize the onboard IMU's measured information to complete the required control state, while simultaneously suppressing drift caused by indirect estimation and integration processes, becomes a problem that requires further investigation. Summary of the Invention

[0006] To address the aforementioned technical problems, the main objective of this invention is to provide a method for estimating the state of an onboard suspension by fusing zero-mean virtual observations. This method enables real-time estimation of the states required for subsequent control, using only the vehicle's existing onboard IMU as the actual measurement sensor, combined with a suspension dynamics model. By improving the accuracy and stability of suspension state estimation under single-sensor conditions, reliable state information can be provided for subsequent semi-active suspension control.

[0007] Another objective of this invention is to propose an onboard suspension state estimation system that integrates zero-mean virtual observations.

[0008] The third objective of this invention is to provide an electronic device.

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

[0010] To achieve the above objectives, a first aspect of the present invention proposes a method for estimating the state of a vehicle suspension by fusing zero-mean virtual observations, comprising:

[0011] A vehicle suspension dynamics model considering road surface roughness excitation, vehicle lateral acceleration and longitudinal acceleration is established, and observation equations are established based on the vehicle vertical acceleration, vehicle pitch angular velocity and vehicle roll angular velocity that can be directly measured by the on-board inertial measurement unit. By utilizing the quasi-zero mean characteristic that the long-term statistical expected mean of vehicle displacement and relative suspension displacement approaches zero, a zero-mean virtual observation is constructed. The zero-mean virtual observation is introduced into the Kalman filter update process, and the state prediction and correction are performed by combining the actual observation and the virtual observation, and the updated posterior suspension state quantity and posterior error covariance matrix are output. The state variables required for suspension control are estimated in real time based on the updated posterior suspension state variables and the posterior error covariance matrix using a recursive Kalman filter algorithm.

[0012] In one embodiment of the present invention, the step of establishing a vehicle suspension dynamics model considering the excitation of road surface roughness, lateral acceleration and longitudinal acceleration of the vehicle body, and establishing observation equations based on the vehicle body vertical acceleration, vehicle body pitch angular velocity and vehicle body roll angular velocity that can be directly measured by the on-board inertial measurement unit includes: Based on Newton's second law, a dynamic differential equation for the whole vehicle suspension system is established under the excitation of road surface roughness and the action of lateral and longitudinal acceleration of the vehicle body. The state vector is defined as the vertical displacement of the vehicle body, pitch angle, roll angle, relative displacement of the four suspensions, vertical velocity of the vehicle body, pitch angular velocity, roll angular velocity, and relative velocity of the four suspensions. The input vector is the semi-active control force of the four suspensions. The system excitation is the longitudinal acceleration, lateral acceleration, and road surface displacement input of the four wheels. The system dynamic equation is written in state-space form. The vehicle vertical acceleration, vehicle pitch angular velocity, and vehicle roll angular velocity, which are directly measured by the vehicle-mounted inertial measurement unit, are selected as actual observations. Virtual observations are constructed by combining the quasi-zero mean characteristics of the vehicle displacement and the relative displacement of the suspension, and the output equation is obtained.

[0013] In one embodiment of the present invention, the step of constructing a zero-mean virtual observation by utilizing the quasi-zero mean characteristic that the long-term statistical expected mean of vehicle body displacement and relative suspension displacement approaches zero includes: The expected values ​​of five state variables, namely the vertical displacement of the vehicle body and the relative displacement of the four suspensions, are set to zero, and a virtual observation containing five zero-value elements is constructed. The virtual observations, together with the vehicle body vertical acceleration, vehicle body pitch rate, and vehicle body roll rate directly measured by the on-board inertial measurement unit, constitute the system observations.

[0014] In one embodiment of the present invention, the step of introducing the zero-mean virtual observation into the Kalman filter update process and combining the actual observation and the virtual observation for state prediction and correction includes: Initialize the system state and error covariance matrix; Predict the prior state vector and the state error covariance matrix based on the discrete state space equation; The Kalman gain matrix is ​​calculated based on the predicted error covariance matrix and the measurement noise covariance matrix. The prior state vector is corrected and updated using the Kalman gain matrix, actual observations, and virtual observations to obtain the estimated value of the posterior optimal state quantity; the posterior state error covariance matrix is ​​then updated.

[0015] In one embodiment of the present invention, the real-time estimation of the state variables required for suspension control based on the updated posterior suspension state variables and the posterior error covariance matrix using a recursive Kalman filter algorithm includes: The posterior optimal state quantity estimate and posterior error covariance matrix estimate obtained after correction and update are substituted into the prior state vector prediction and state error covariance matrix prediction steps at the next time step to realize the Kalman filter cyclic recursive process, thereby outputting suspension state information in real time, including vehicle vertical displacement, pitch angle, roll angle, relative displacement of the four suspensions, vehicle vertical velocity, pitch angular velocity, roll angular velocity, and relative velocity of the four suspensions.

[0016] To achieve the above objectives, a second aspect of the present invention proposes a vehicle suspension state estimation system that integrates zero-mean virtual observations, comprising: The observation equation construction module is used to establish a vehicle suspension dynamics model that considers the excitation of road surface roughness, the lateral acceleration and longitudinal acceleration of the vehicle body, and to establish observation equations based on the vehicle body vertical acceleration, vehicle body pitch angular velocity and vehicle body roll angular velocity that can be directly measured by the on-board inertial measurement unit. The observation construction module is used to construct zero-mean virtual observations by taking advantage of the near-zero mean characteristic that the long-term statistical expected mean of vehicle body displacement and relative suspension displacement approaches zero. The prediction and correction module is used to introduce the zero-mean virtual observation into the Kalman filter update process, combine the actual observation and the virtual observation to perform state prediction and correction, and output the updated posterior suspension state quantity and posterior error covariance matrix. The state quantity estimation module is used to estimate the state quantities required for suspension control in real time based on the updated posterior suspension state quantities and the posterior error covariance matrix using a recursive Kalman filter algorithm.

[0017] The vehicle suspension state estimation method and system based on zero-mean virtual observations, as described in this invention, uses the vehicle-mounted IMU sensors as the actual measurement source, without relying on other vehicle state measurement sensors. First, a whole-vehicle suspension dynamics model is established, considering road roughness excitation, lateral acceleration, and longitudinal acceleration of the vehicle body. Observation equations are then established based on the vehicle body vertical acceleration, pitch angular velocity, and roll angular velocity, which can be directly measured by the onboard IMU. On this basis, utilizing the quasi-zero-mean characteristic that the long-term statistical expected mean of vehicle body displacement and relative suspension displacement approaches zero, zero-mean virtual observations are constructed and an update process using Kalman filtering is introduced to estimate the state variables required for subsequent suspension control in real time. Through simulation and prototype vehicle testing, this invention can suppress integral drift and improve the accuracy and stability of suspension state estimation by relying solely on the actual measurement signals from the onboard IMU, providing reliable state information for subsequent suspension control.

[0018] A third objective of this invention is to provide an electronic device comprising: processor; The memory stores executable instructions; when the processor executes the instructions, it implements the vehicle suspension state estimation method fused with zero-mean virtual observations as described above.

[0019] The fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for estimating the state of a vehicle suspension by fusing zero-mean virtual observations.

[0020] 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

[0021] 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: Figure 1 A flowchart illustrating a method for estimating the state of a vehicle suspension by fusing zero-mean virtual observations, provided in an embodiment of the present invention; Figure 2 A diagram of a vehicle suspension dynamics model provided in an embodiment of the present invention, based on road surface roughness excitation and the lateral and longitudinal accelerations of the vehicle body. Figure 3 This is a schematic diagram showing the position of the sprung center of gravity from the vehicle's center of gravity, provided in an embodiment of the present invention. Figure 4 A comparison chart of suspension state estimation results under linear deceleration conditions provided in an embodiment of the present invention; Figure 5 An estimated state quantity diagram for a dual-track shifting operation provided in an embodiment of the present invention; Figure 6 This is a comparison chart of suspension state estimation results under dual lane change conditions provided in an embodiment of the present invention; Figure 7 This is a diagram of a curve simulation test scenario provided in an embodiment of the present invention; Figure 8 Comparison chart of suspension state estimation results under cornering deceleration conditions provided in embodiments of the present invention; Figure 9 A diagram showing the suspension dynamic deflection estimation under cornering deceleration conditions provided in an embodiment of the present invention; Figure 10 This is a structural diagram of a vehicle suspension state estimation system that integrates zero-mean virtual observations, provided in an embodiment of the present invention. Figure 11 The computer device provided in the embodiments of the present invention. Detailed Implementation

[0022] 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.

[0023] 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.

[0024] The following description, with reference to the accompanying drawings, describes a method, system, computer device, and storage medium for estimating the state of a vehicle suspension system based on zero-mean virtual observations, according to an embodiment of the present invention.

[0025] Example 1 This embodiment provides a method for estimating the state of a vehicle suspension by fusing zero-mean virtual observations. For example... Figure 1 As shown, it includes: S1. Establish a vehicle suspension dynamics model that considers the excitation of road surface roughness, the effects of vehicle body lateral acceleration and longitudinal acceleration, and establish observation equations based on the vehicle body vertical acceleration, vehicle body pitch angular velocity and vehicle body roll angular velocity that can be directly measured by the on-board inertial measurement unit.

[0026] Specifically, firstly, a vehicle suspension dynamics model is constructed. This model must comprehensively consider the effects of road surface roughness excitation and the lateral and longitudinal accelerations of the vehicle body on the dynamic behavior of the suspension system. Based on Newtonian mechanics, the model treats the vehicle body as a rigid body with vertical, pitch, and roll degrees of freedom, and describes the relative motion at each suspension point, thus establishing a state-space expression that includes the vertical displacement of the vehicle's center of mass, pitch angle, roll angle, relative displacements of each suspension point, and their corresponding velocities. Building upon this, based on the physical quantities that the onboard inertial measurement unit can directly measure—namely, the vehicle's vertical acceleration, pitch velocity, and roll velocity—observation equations are established. These equations express the mapping relationship between the system state vector and these measurable output quantities in linear or nonlinear form. These observation equations serve as the source of actual measurement information in subsequent filtering estimations, ensuring the observability of the system state.

[0027] By establishing the vehicle suspension dynamics model that includes road excitation and vehicle body lateral and longitudinal acceleration, and constructing observation equations based on the directly measurable signals from the onboard IMU, an accurate system dynamic description and reliable measurement information basis are provided for subsequent state estimation. This enables the effective observation of the state quantities required for suspension control to be achieved by relying on only a single onboard IMU sensor, thereby reducing the cost and layout complexity caused by multi-sensor configuration.

[0028] S2 utilizes the near-zero mean characteristic that the long-term statistical expected mean of vehicle body displacement and relative suspension displacement approaches zero to construct a zero-mean virtual observation.

[0029] Specifically, zero-mean virtual observations are constructed by leveraging the near-zero-mean characteristic that the long-term statistical expected mean of vehicle body displacement and relative suspension displacement approaches zero. Specifically, based on the physical characteristic that the statistical mean of the vertical displacement of the vehicle body and the relative displacement at each suspension point tends to zero over a long time scale during normal vehicle operation, these state variables are introduced into the observation equation as virtual observations, with their observed values ​​set to zero. This provides additional constraint information for Kalman filtering without adding extra physical sensors. As one implementation method, five state variables—the vertical displacement of the vehicle body and the four relative displacements of the four suspensions—can be used as virtual observations, all with zero observed values, and corresponding measurement noise covariance is assigned to adjust their weights in the filter update. In this way, virtual observations can suppress integral drift in state estimation, enabling the filtering algorithm to maintain accurate tracking of suspension state variables even when relying solely on the actual measurement signals from the onboard inertial measurement unit.

[0030] By introducing zero-mean virtual observations based on quasi-zero-mean characteristics, the cumulative error and drift caused by acceleration integration can be effectively suppressed without increasing hardware costs, thereby improving the long-term stability and accuracy of suspension state estimation and providing reliable state information for subsequent suspension control.

[0031] S3, the zero-mean virtual observation is introduced into the Kalman filter update process, and the state prediction and correction are performed by combining the actual observation and the virtual observation, and the updated posterior suspension state quantity and posterior error covariance matrix are output.

[0032] Specifically, after obtaining zero-mean virtual observations, they are combined with actual observations directly measured by the onboard inertial measurement unit and introduced into the Kalman filter framework to form an update mechanism that integrates virtual and measured information. Specifically, in the prediction phase of the Kalman filter, a priori estimation of the system state is performed based on a discretized vehicle suspension dynamics model, yielding the predicted state value and its error covariance matrix. In the update phase, both actual and zero-mean virtual observations are used to correct the prior estimate. The actual observations directly reflect the vehicle's motion state at the current moment, while the zero-mean virtual observations, based on the quasi-zero-mean characteristic that the long-term statistical expectation of vehicle displacement and suspension relative displacement approaches zero, provide additional constraint information for state updates. By combining these two types of observations into a unified observation vector and setting the corresponding measurement noise covariance matrix according to the uncertainty of each observation channel, the Kalman gain can adaptively balance the credibility between the predicted value and different observation information, thereby achieving optimal fusion of the posterior state estimate.

[0033] As one implementation method, the vehicle vertical acceleration, pitch angular velocity, and roll angular velocity measured by the on-board IMU can be combined with five zero-mean virtual observations (vehicle vertical displacement and four suspension relative displacements) to form an eight-dimensional observation vector. Based on the statistical characteristics of IMU sensor noise and the high uncertainty of virtual observations, the diagonal elements of the measurement noise covariance matrix are set respectively, and then the state prediction and correction update are completed according to the standard Kalman filter recursive formula.

[0034] S4 uses a recursive Kalman filter algorithm to estimate the state variables required for suspension control in real time based on the updated posterior suspension state variables and the posterior error covariance matrix.

[0035] Specifically, after completing the prediction and correction update of the Kalman filter, a recursive algorithm architecture is used to take the posterior state vector and its error covariance matrix obtained from the correction update at the current time as the initial input for the prediction process at the next time, thus forming a continuously recursive closed loop for state estimation. Specifically, this method uses the discretized system state-space equations to predict the prior state at the current time based on the posterior state estimate from the previous time step and the known system inputs (including control input and excitation input), and simultaneously predicts the prior error covariance matrix. Subsequently, based on the observation vector composed of actual and virtual observations, the Kalman gain is calculated by combining the predicted prior state and the observation equation. This gain is used to balance the reliability of the prediction information and the observation information. Then, the Kalman gain is used to correct the prior state, obtaining the optimal estimate of the posterior state at the current time step, and updating the posterior error covariance matrix. The above prediction and correction process is repeated in each sampling period, enabling the state estimation to be continuously updated with the dynamic changes of the system and new observation data, thereby achieving real-time recursive estimation of the state quantities required for suspension control. As one implementation method, the recursive process may specifically include: initializing the state vector and error covariance matrix; performing prior state prediction and prior covariance prediction based on the discrete state-space model; calculating the Kalman gain; correcting and updating the prior state using actual and virtual observations; updating the posterior covariance matrix; and substituting the updated posterior state and covariance matrix into the prediction step at the next time step, thus iterating in this way to continuously output the suspension state estimate.

[0036] The aforementioned recursive Kalman filter algorithm can continuously fuse actual and virtual observations in the time dimension. By utilizing the suppression effect of virtual observations on integral drift, the algorithm accumulates continuously during the recursive process, thereby effectively suppressing state estimation drift caused by acceleration integral and noise accumulation. This improves the long-term stability and real-time tracking accuracy of suspension state estimation, providing reliable and continuous state information for subsequent suspension control.

[0037] Example 2 This embodiment addresses the issue that suspension control systems in practical vehicle applications require control state information such as vehicle posture, suspension dynamic deflection, and their changing trends. Multi-sensor measurement schemes suffer from problems such as high sensor cost, complex arrangement, and significant measurement noise. Furthermore, using only the vehicle's existing onboard IMU as a single actual measurement sensor, supplementing the required control state through acceleration or velocity signal integration easily leads to accumulated errors and drift. This invention does not consider wheel dynamics characteristics and establishes a whole-vehicle suspension dynamics model based on road surface unevenness excitation and the effects of lateral and longitudinal vehicle acceleration, as follows: Figure 2 As shown: in, , , , 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 left front, right front, left rear, and right rear tires, respectively. , , , These represent the vertical displacements of the vehicle body at the front left, front right, rear left, and rear right, respectively. , These represent the damping coefficients of the shock absorbers for individual suspensions on the front and rear axles, respectively. , These represent the spring stiffness of individual suspension components on the front and rear axles, respectively. , These represent the tire stiffness of a single wheel on the front and rear axles, respectively. For the sprung mass, For unsprung mass, , These are the moments of inertia for vehicle pitch and roll, respectively. This represents the vertical displacement of the vehicle's center of gravity. This refers to the vehicle body pitch angle displacement. This refers to the vehicle body roll angle displacement. , These are the distances from the vehicle's center of gravity to the front and rear axles, respectively. It is half the wheel track. , These represent longitudinal acceleration and lateral acceleration, respectively.

[0038] Based on Newton's second law, this implementation considers the vertical, pitch, and roll motions of the vehicle body caused by road surface excitation and lateral and longitudinal accelerations, and establishes the dynamic differential equations of the whole vehicle suspension system model as shown in (1):

[0039] In the formula, The distance from the sprung center of mass to the vehicle's center of mass, such as... Figure 3 As shown, for example, , , , The table shows the dynamic forces of the left front, right front, left rear, and right rear suspensions, and their calculation formulas are shown in (2):

[0040] At small pitch angles and small roll angle Within this range, it can be approximated as:

[0041] Therefore, the absolute displacement of the vehicle body at each suspension point can be approximated as:

[0042] Define state vector Input vectors are used to measure the vehicle's vertical displacement, pitch angle, roll angle, relative suspension displacement, and the vehicle's vertical velocity, pitch angular velocity, roll angular velocity, and relative suspension velocity. For the semi-active control force of the four suspensions, the system excitation Inputs include longitudinal and lateral accelerations, as well as four-wheel road surface displacements, i.e.:

[0043] Combining equations (2)-(4), the system dynamic equation (1) can be written in state-space form:

[0044] In the formula, the coefficient matrix , , The expression is as follows: ; ; ; ; .

[0045] ; ; .

[0046] ; ; ; ; ; ; ; .

[0047] Furthermore, the vehicle's vertical acceleration, pitch velocity, and roll velocity, directly measured by the onboard IMU, are selected as actual observations. Virtual observations are constructed considering the quasi-zero mean characteristics of the vehicle displacement and the relative displacement of the suspension. The final system observations are expressed as follows:

[0048] Combining equation (6), the output equation can be obtained as follows:

[0049] In the formula, the matrix The expression is: ； ； ； ； ； ； 。

[0050] matrix The expression is: ， 。

[0051] From equations (8) and (2), the state-space equation for the fused zero-mean virtual observations can be expressed as:

[0052] In the formula, For process noise, To measure the noise, both are assumed to be zero-mean Gaussian white noise and to be independent of each other.

[0053] Based on this, discretizing the system yields:

[0054] In the formula, , , , , , The sampling time interval is set to 0.01 s, and the noise covariance matrix is: , .

[0055] According to modern control theory, complete state observability is a prerequisite for uniquely reconstructing state variables solely from the system's output sequence. For linear time-invariant systems, observability is often verified using the rank criterion of the observability matrix, i.e.:

[0056] Upon inspection, 14. The observability condition is satisfied. Therefore, all state variables in the constructed discrete state space can be observed.

[0057] In summary, the Kalman filter suspension state estimation algorithm that integrates zero-mean virtual observations can be divided into two processes: prediction and update. The steps are as follows: Step 1: Initialize the system state and error covariance matrix:

[0058] Step 2: Prediction of prior state vectors:

[0059] In the formula, for The optimal estimate of the posterior state vector at time step [time]. for The actual control force of the semi-active suspension at any given moment. for Always The predicted value of the prior state vector at time step.

[0060] Step 3: Prediction of the state error covariance matrix:

[0061] In the formula, The process noise covariance matrix is... for The estimated value of the posterior error covariance matrix at time step [time]. for Always The predicted value of the prior error covariance matrix at time step.

[0062] Step 4: Calculate the Kalman gain:

[0063] In the formula, express The Kalman gain matrix at time 1 is used to quantitatively weigh the relative reliability of the predicted and measured values. The measurement noise error covariance matrix is ​​represented by its diagonal elements, which characterize the noise variance of the corresponding measurement channel. A larger value indicates lower reliability of the corresponding measurement information. The IMU sensor used in this invention is the LGA-14L type, and its vertical acceleration, pitch angular velocity, and roll angular velocity noise statistical characteristics are shown in Table 1. For zero-mean virtual observations, due to the lack of direct physical measurement support, the uncertainty is significantly higher than that of the IMU measured signal; therefore, a larger value is assigned... The value (0.0005) is set to reduce its weight in state updates, ensuring that the filtered estimation relies primarily on reliable sensor measurements.

[0064] Table 1: Measurement Noise Error

[0065] Step 5: Correct and update the posterior state vector:

[0066] In the formula, for The estimated values ​​of the posterior optimal state variables at time step 1 are obtained by weighted fusion of the prior state vectors. and observation information get.

[0067] Step 6: Update the posterior state error covariance matrix:

[0068] In the formula, express The estimated value of the posterior error covariance matrix at time step 1 is used for prediction at the next time step.

[0069] The updated and Substitute (13) and (14) into the loop recursive process of ZVO-KF to estimate the suspension state information in real time.

[0070] Example 3 To verify the performance of the Kalman filter observer with fused zero-mean virtual observations (ZVO-KF) for state estimation of the suspension system, a CarSim / Simulink co-simulation platform was built, and the sprung acceleration at the center of mass was extracted from the CarSim model output. Vehicle pitch rate Vehicle pitch rate The actual measured IMU noise is superimposed to simulate the observations collected by the IMU during the actual driving process of the vehicle; three driving scenarios are constructed respectively: straight deceleration, double lane change and cornering deceleration, and the simulation is compared with the traditional Kalman filter (KF).

[0071] (1) Linear deceleration condition: Using only IMU measurement signals, the actual suspension state variables under linear deceleration conditions are compared with the ZVO-KF and KF estimated state variables as follows: Figure 4 As shown. Figure 4 In the figure, (a) represents the relative displacement of the suspension. Figure 4 (b) represents the relative speed of the suspension. The quantitative results are shown in Table 2.

[0072] Depend on Figure 4 It can be seen that within the first 12 seconds, both methods accurately estimated the relative displacement and relative velocity of each suspension. After 12 seconds, the traditional KF method, affected by IMU measurement noise, accumulated errors in estimating the relative velocities of the front and rear suspensions. The integral of the suspension relative velocity error further amplified the drift in relative displacement. In contrast, the ZVO-KF method, by introducing five zero-mean virtual observations—the vertical displacement of the vehicle body and the relative displacements of the four suspensions—effectively corrected the integral accumulation error and suppressed the integral drift of relative displacement and relative velocity. The quantitative analysis results in Table 2 show that the ZVO-KF method has a much higher accuracy in estimating the relative displacement and relative velocity of the suspension than the traditional KF method.

[0073] Table 2: Prediction accuracy of suspension condition under cornering deceleration conditions

[0074] (2) Double line shifting operation: In CarSim, a vehicle is set to undergo a double lane change test at a constant speed of 70 km / h. The simulated test conditions are as follows: Figure 5 As shown. The actual suspension state variables and the ZVO-KF and KF estimated state variables under dual lane change conditions, using only IMU measurement signals, are as follows: Figure 6 As shown, Figure 6 In the figure, (a) represents the relative displacement of the suspension. Figure 6 (b) represents the relative speed of the suspension. The quantitative analysis results are shown in Table 3.

[0075] Table 3: Suspension State Estimation Accuracy under Dual Lane Change Conditions

[0076] Depend on Figure 6It can be seen that after 8.5 seconds, the traditional KF method is affected by IMU measurement noise, resulting in cumulative errors in the estimation of relative velocities of each suspension. The integral of the suspension relative velocity error further amplifies the drift of relative displacement. Because the amplitude of the vehicle's dynamic response is significantly higher than the noise interference level under extreme conditions such as double lane change, the degree of suspension drift is reduced compared to straight-line deceleration conditions. Table 3 shows that the ZVO-KF method outperforms the traditional KF method in estimating the relative displacement and relative velocity of the suspension.

[0077] (3) Deceleration during curves: Building in CarSim Figure 7 The illustrated "straight-curve-straight" scenario uses a spiral connection between straight sections and curves to ensure a continuous transition in road curvature. The curve radius is set according to GB 55011-2021 standard. The vehicle speed is set to a constant 108 km / h from 0 to 5 seconds, then a uniform deceleration from 108 km / h to 72 km / h from 5 to 10 seconds, and finally a constant 72 km / h after 10 seconds. This road scenario and speed planning ensure that the vehicle passes through four typical driving conditions: constant speed on a straight road, deceleration on a straight road, deceleration on a curve, and constant speed on a curve.

[0078] Set the vehicle in such Figure 7 The actual suspension state variables under the curve deceleration condition, measured only by IMU signals, are compared with the estimated state variables of ZVO-KF and KF as shown in the diagram. Figure 8 As shown, Figure 8 In the figure, (a) represents the relative displacement of the suspension. Figure 8 (b) represents the relative speed of the suspension. The quantitative analysis results are shown in Table 4.

[0079] Table 4: Suspension State Estimation Accuracy under Cornering Deceleration Condition

[0080] Depend on Figure 8 It is evident that the traditional KF method, relying on IMU measurement information for state estimation, lacks other measurement information to correct for errors caused by integration accumulation. This leads to continuous noise accumulation during acceleration integration, resulting in significant drift in the estimation results of suspension relative displacement and relative velocity, with the estimation error diverging over time. In contrast, the ZVO-KF method, by introducing zero-mean virtual observations, effectively corrects for integration accumulation errors and suppresses integration drift. Table 4 shows that the quantitative analysis results indicate that the ZVO-KF method has significantly higher accuracy in estimating the suspension relative displacement and relative velocity state than the traditional KF method.

[0081] Example 4 To verify the effectiveness of the ZVO-KF state estimation algorithm in a real vehicle environment, a typical driving condition involving cornering deceleration was selected for real vehicle testing. The test conditions were set as follows: (1) Distribution of vehicle testing personnel: Driver: The driver is a professional tester to ensure the safety of the driver and passengers during the test and the effectiveness of each test condition.

[0082] Tester 1: Sitting behind the driver's seat, collecting the vehicle's vertical acceleration, pitch rate, roll rate from the IMU, and the suspension state data estimated by the ZVO-KF and the control output from the controller.

[0083] Tester 2: Sitting behind the passenger seat, recording the vehicle's CAN data using the ZLG USBCANFD-200U bus analyzer.

[0084] (2) Test conditions: Curve deceleration condition: Maintain a constant speed of 70 km / h on the straight road. Before entering the left U-shaped curve, the driver decelerates with a fixed brake pedal opening. Continue to decelerate to 50 km / h inside the curve and then maintain a constant speed.

[0085] Depend on Figure 9 It can be seen that at approximately 5 seconds, the test vehicle began braking and entered the curve. Under the combined action of braking and steering, the vehicle experienced longitudinal and lateral load transfer, and the dynamic deflection at all four suspension points showed significant changes. Comparing the CAN data collected from the actual vehicle with the ZVO-KF estimated values, it can be seen that the ZVO-KF estimation results can reflect the changing trend of suspension dynamic deflection during the deceleration process in the curve. No obvious integral drift phenomenon was observed during the braking entry into the curve and the driving phase in the curve, indicating that the zero-mean virtual observation has a certain suppressive effect on the estimated drift.

[0086] Meanwhile, a certain amplitude deviation still exists between the ZVO-KF estimated values ​​and the values ​​collected by the actual vehicle's CAN bus. This deviation is mainly related to the nonlinearity of the estimation model parameters and the actual suspension system. Under cornering deceleration conditions, the vehicle experiences both longitudinal load transfer caused by braking and lateral load transfer caused by steering. The estimation model uses linear parameters to describe the suspension's dynamic characteristics, making it difficult to fully reflect the impact of actual load changes on the suspension's dynamic deflection. Therefore, the estimated results differ in amplitude from the CAN bus's collected values. In summary, although the ZVO-KF estimation algorithm has some estimation bias, it can reflect the response trend of the suspension's dynamic deflection relatively well and maintains good numerical stability during real-vehicle testing, demonstrating its feasibility for engineering applications.

[0087] Example 5 To implement the methods of the above embodiments, the present invention also provides a vehicle suspension state estimation system 10 that integrates zero-mean virtual observations, such as... Figure 10 As shown, it includes: The observation equation construction module 100 is used to establish a vehicle suspension dynamics model that considers the excitation of road surface roughness, the lateral acceleration and longitudinal acceleration of the vehicle body, and to establish observation equations based on the vehicle body vertical acceleration, vehicle body pitch angular velocity and vehicle body roll angular velocity that can be directly measured by the on-board inertial measurement unit. The observation construction module 200 is used to construct zero-mean virtual observations by utilizing the quasi-zero mean characteristic that the long-term statistical expected mean of vehicle displacement and relative suspension displacement approaches zero. The prediction and correction module 300 is used to introduce the zero-mean virtual observation into the Kalman filter update process, combine the actual observation and the virtual observation to perform state prediction and correction, and output the updated posterior suspension state quantity and posterior error covariance matrix. The state quantity estimation module 400 is used to estimate the state quantities required for suspension control in real time based on the updated posterior suspension state quantities and the posterior error covariance matrix using a recursive Kalman filter algorithm.

[0088] Furthermore, the observation equation construction module 100 is also used for: Based on Newton's second law, a dynamic differential equation for the whole vehicle suspension system is established under the excitation of road surface roughness and the action of lateral and longitudinal acceleration of the vehicle body. The state vector is defined as the vertical displacement of the vehicle body, pitch angle, roll angle, relative displacement of the four suspensions, vertical velocity of the vehicle body, pitch angular velocity, roll angular velocity, and relative velocity of the four suspensions. The input vector is the semi-active control force of the four suspensions. The system excitation is the longitudinal acceleration, lateral acceleration, and road surface displacement input of the four wheels. The system dynamic equation is written in state-space form. The vehicle vertical acceleration, vehicle pitch angular velocity, and vehicle roll angular velocity, which are directly measured by the vehicle-mounted inertial measurement unit, are selected as actual observations. Virtual observations are constructed by combining the quasi-zero mean characteristics of the vehicle displacement and the relative displacement of the suspension, and the output equation is obtained.

[0089] Furthermore, the observation construction module is also used for: The expected values ​​of five state variables, namely the vertical displacement of the vehicle body and the relative displacement of the four suspensions, are set to zero, and a virtual observation containing five zero-value elements is constructed. The virtual observations, together with the vehicle body vertical acceleration, vehicle body pitch rate, and vehicle body roll rate directly measured by the on-board inertial measurement unit, constitute the system observations.

[0090] The vehicle suspension state estimation system based on zero-mean virtual observations in this invention can suppress integral drift and improve the accuracy and stability of suspension state estimation by relying solely on the actual measurement signals from the vehicle IMU, thus providing reliable state information for subsequent suspension control.

[0091] Example 6 To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 11 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads 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 method described above.

[0092] Example 7 To implement the methods of the above embodiments, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, is used to implement the various steps of the methods described above.

[0093] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0094] 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.

[0095] 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. A method for estimating the state of a vehicle suspension by fusing zero-mean virtual observations, characterized in that, include: A vehicle suspension dynamics model considering road surface roughness excitation, vehicle lateral acceleration and longitudinal acceleration is established, and observation equations are established based on the vehicle vertical acceleration, vehicle pitch angular velocity and vehicle roll angular velocity that can be directly measured by the on-board inertial measurement unit. By utilizing the quasi-zero mean characteristic that the long-term statistical expected mean of vehicle displacement and relative suspension displacement approaches zero, a zero-mean virtual observation is constructed. The zero-mean virtual observation is introduced into the Kalman filter update process, and the state prediction and correction are performed by combining the actual observation and the virtual observation, and the updated posterior suspension state quantity and posterior error covariance matrix are output. The state variables required for suspension control are estimated in real time based on the updated posterior suspension state variables and the posterior error covariance matrix using a recursive Kalman filter algorithm.

2. The method as described in claim 1, characterized in that, The establishment of a vehicle suspension dynamics model considering road surface roughness excitation, lateral acceleration, and longitudinal acceleration of the vehicle body, and the establishment of observation equations based on the vehicle body vertical acceleration, vehicle body pitch angular velocity, and vehicle body roll angular velocity that can be directly measured by the onboard inertial measurement unit, includes: Based on Newton's second law, a dynamic differential equation for the whole vehicle suspension system is established under the excitation of road surface roughness and the action of lateral and longitudinal acceleration of the vehicle body. The state vector is defined as the vertical displacement of the vehicle body, pitch angle, roll angle, relative displacement of the four suspensions, vertical velocity of the vehicle body, pitch angular velocity, roll angular velocity, and relative velocity of the four suspensions. The input vector is the semi-active control force of the four suspensions. The system excitation is the longitudinal acceleration, lateral acceleration, and road surface displacement input of the four wheels. The system dynamic equation is written in state-space form. The vehicle vertical acceleration, vehicle pitch angular velocity, and vehicle roll angular velocity, which are directly measured by the vehicle-mounted inertial measurement unit, are selected as actual observations. Virtual observations are constructed by combining the quasi-zero mean characteristics of the vehicle displacement and the relative displacement of the suspension, and the output equation is obtained.

3. The method as described in claim 1, characterized in that, The method of constructing a zero-mean virtual observation by utilizing the quasi-zero mean characteristic that the long-term statistical expected mean of vehicle body displacement and relative suspension displacement approaches zero includes: The expected values ​​of five state variables, namely the vertical displacement of the vehicle body and the relative displacement of the four suspensions, are set to zero, and a virtual observation containing five zero-value elements is constructed. The virtual observations, together with the vehicle body vertical acceleration, vehicle body pitch rate, and vehicle body roll rate directly measured by the on-board inertial measurement unit, constitute the system observations.

4. The method as described in claim 1, characterized in that, The step of incorporating the zero-mean virtual observation into the Kalman filter update process and combining the actual observation and the virtual observation for state prediction and correction includes: Initialize the system state and error covariance matrix; Predict the prior state vector and the state error covariance matrix based on the discrete state space equation; The Kalman gain matrix is ​​calculated based on the predicted error covariance matrix and the measurement noise covariance matrix. The prior state vector is corrected and updated using the Kalman gain matrix, actual observations, and virtual observations to obtain the estimated value of the posterior optimal state quantity; the posterior state error covariance matrix is ​​then updated.

5. The method as described in claim 1, characterized in that, The Kalman filter algorithm, which uses a cyclic recursion to estimate the state variables required for suspension control in real time based on the updated posterior suspension state variables and the posterior error covariance matrix, includes: The posterior optimal state quantity estimate and posterior error covariance matrix estimate obtained after correction and update are substituted into the prior state vector prediction and state error covariance matrix prediction steps at the next time step to realize the Kalman filter cyclic recursive process, thereby outputting suspension state information in real time, including vehicle vertical displacement, pitch angle, roll angle, relative displacement of the four suspensions, vehicle vertical velocity, pitch angular velocity, roll angular velocity, and relative velocity of the four suspensions.

6. A vehicle suspension state estimation system integrating zero-mean virtual observations, characterized in that, include: The observation equation construction module is used to establish a vehicle suspension dynamics model that considers the excitation of road surface roughness, the lateral acceleration and longitudinal acceleration of the vehicle body, and to establish observation equations based on the vehicle body vertical acceleration, vehicle body pitch angular velocity and vehicle body roll angular velocity that can be directly measured by the on-board inertial measurement unit. The observation construction module is used to construct zero-mean virtual observations by taking advantage of the near-zero mean characteristic that the long-term statistical expected mean of vehicle body displacement and relative suspension displacement approaches zero. The prediction and correction module is used to introduce the zero-mean virtual observation into the Kalman filter update process, combine the actual observation and the virtual observation to perform state prediction and correction, and output the updated posterior suspension state quantity and posterior error covariance matrix. The state quantity estimation module is used to estimate the state quantities required for suspension control in real time based on the updated posterior suspension state quantities and the posterior error covariance matrix using a recursive Kalman filter algorithm.

7. The system as described in claim 6, characterized in that, The observation equation construction module is also used for: Based on Newton's second law, a dynamic differential equation for the whole vehicle suspension system is established under the excitation of road surface roughness and the action of lateral and longitudinal acceleration of the vehicle body. The state vector is defined as the vertical displacement of the vehicle body, pitch angle, roll angle, relative displacement of the four suspensions, vertical velocity of the vehicle body, pitch angular velocity, roll angular velocity, and relative velocity of the four suspensions. The input vector is the semi-active control force of the four suspensions. The system excitation is the longitudinal acceleration, lateral acceleration, and road surface displacement input of the four wheels. The system dynamic equation is written in state-space form. The vehicle vertical acceleration, vehicle pitch angular velocity, and vehicle roll angular velocity, which are directly measured by the vehicle-mounted inertial measurement unit, are selected as actual observations. Virtual observations are constructed by combining the quasi-zero mean characteristics of the vehicle displacement and the relative displacement of the suspension, and the output equation is obtained.

8. The system as described in claim 6, characterized in that, The observation construction module is also used for: The expected values ​​of five state variables, namely the vertical displacement of the vehicle body and the relative displacement of the four suspensions, are set to zero, and a virtual observation containing five zero-value elements is constructed. The virtual observations, together with the vehicle body vertical acceleration, vehicle body pitch rate, and vehicle body roll rate directly measured by the on-board inertial measurement unit, constitute the system observations.

9. An electronic device, comprising: processor; The memory stores executable instructions; when the processor executes the instructions, it implements the method as described in any one of claims 1-5.

10. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method as claimed in any one of claims 1-5.

Images