A method for estimating longitudinal speed of a four-wheel drive hub motor vehicle

Abstract

The present invention provides a method for estimating the longitudinal speed of a four-wheel drive hub motor vehicle, ensuring that the longitudinal speed can be estimated stably and accurately under complex nonlinear working conditions where the wheels slip or lock. The method first collects vehicle state parameters, uses an extended sensor data method, improves the volumetric Kalman filter method, adds a balance weighting factor, and obtains the longitudinal speed observation value through weighted fusion; secondly, four driving wheel speed signals are used for comprehensive discrimination, the current driving state of the vehicle is identified in real time, fuzzy logic rules are formulated to obtain the similarity coefficient of each driving wheel, and the longitudinal speed estimation value is obtained through fusion integration; finally, a comprehensive analysis is performed based on the longitudinal speed observation value and the estimated value, and the final longitudinal speed is obtained through weighted fusion by introducing a working condition adaptation factor. The method has good adaptability and strong stability in a nonlinear strongly coupled system, and improves the robustness of the longitudinal speed estimation under complex working conditions.

Description

Technical Field

[0001] The invention relates to an electric vehicle driven by a hub motor, and in particular to a method for estimating the longitudinal speed of a four-wheel drive hub motor vehicle, belonging to the field of active safety control of electric vehicles. Background Art

[0002] With the rapid development of electric vehicle technology, the application of hub motors has become mainstream, and the active safety control of hub motor vehicles is a key technology for hub motor vehicles. Among them, the longitudinal speed of the vehicle is an important parameter in the vehicle's active safety control system. For traditional fuel vehicles, the longitudinal speed can be estimated using the wheel speed information of the non-driven wheels, but for four-wheel hub motor electric vehicles, all four wheels are driven wheels, and the wheel speed of the non-driven wheels can no longer be used to estimate the vehicle speed. The algorithms commonly used for vehicle longitudinal speed estimation currently include Kalman filtering, observer-based method, radar / GPS measurement method, and wheel model information fusion method. However, when a wheel of the vehicle slips or locks, the wheel speed will increase or decrease sharply. At this time, the wheel model and state observer are easily disturbed by the outside world, making the stability of the estimation method unsatisfactory.

[0003] Chinese patent application No. 201910885710.X proposes a "vehicle speed estimation method for hub motor driven vehicles based on multi-model fusion". This method fuses different sensor signals, uses the Kalman filter algorithm to filter the data, and realizes vehicle speed estimation by using a weighted fusion method. However, the patented technical solution adopts a linear estimation method, ignoring the coupling during vehicle motion under complex working conditions, and the adaptability of this solution in the vehicle nonlinear motion system is not satisfied. Chinese patent application No. 2017100896997 discloses a "distributed drive electric vehicle road adaptive longitudinal vehicle speed estimation system and method". This invention estimates the longitudinal vehicle speed online based on the interconnection of the longitudinal force estimator and the road peak adhesion coefficient estimator. This method has the characteristics of small calculation amount and strong versatility, but this method does not take into account the influence of the vehicle dynamic characteristics on the sensor collected data during actual driving. When the obtained signal fluctuates greatly, it is easy to cause the speed estimation accuracy to decrease.

[0004] The CKF (Cubic Kalman Filter) is an estimation method based on nonlinear systems. It is a new Gaussian filtering method based on the volumetric numerical integration criterion. Its core is to select multiple volumetric points with equal weights and use the spherical diameter integration criterion to achieve nonlinear estimation. Since the weight of each volumetric point used in the CKF is the same, it can adapt to complex nonlinear strongly coupled systems. However, during the CKF process, its iterative signal is easily interfered by the outside world, especially when the vehicle enters complex working conditions, the filter signal error is large and the estimation result is inaccurate.

[0005] Therefore, in view of the problem that the adaptability, stability and accuracy of the existing longitudinal vehicle speed estimation methods cannot be met under the nonlinear and complex working conditions of four-wheel drive hub motor vehicles slipping or locking, designing a longitudinal vehicle speed estimation method with wide adaptability, strong stability and high accuracy has important theoretical significance and practical value. Summary of the invention

[0006] In view of the above problems, the purpose of the present invention is to provide a method for estimating the longitudinal speed of a four-wheel independent drive hub motor vehicle. The method first collects vehicle state parameters, uses an extended sensor data method, improves the volumetric Kalman filter method, adds a balance weighting factor, and obtains the longitudinal speed observation value through weighted fusion; secondly, the four driving wheel speed signals are used for comprehensive discrimination, the current driving state of the vehicle is identified in real time, fuzzy logic rules are formulated to obtain the similarity coefficient of each driving wheel, and the longitudinal speed estimation value is obtained through fusion integration; finally, a comprehensive analysis is performed based on the longitudinal speed observation value and the estimated value, and the final longitudinal speed is obtained by weighted fusion by introducing the working condition adaptation factor. The method has good adaptability and strong stability under complex nonlinear working conditions, and can realize real-time and accurate longitudinal speed estimation.

[0007] The present invention is achieved by adopting the following technical solutions:

[0008] Step 1: First, collect vehicle state parameters, use the extended sensor data method, then improve the volumetric Kalman filter method, add a balance weighting factor, and finally obtain the longitudinal speed observation value through weighted fusion:

[0009] (1) obtaining vehicle configuration parameters according to vehicle-mounted sensors and collecting vehicle status parameters through a data bus; wherein the vehicle configuration parameters include vehicle mass, vehicle body moment of inertia, vehicle wheelbase and wheel track; the vehicle status parameters include four driving wheel speeds, driving wheel rotation angular velocity, longitudinal acceleration, lateral acceleration, yaw angular velocity and front wheel steering angle; the data bus includes a vehicle CAN bus and a FlexRay bus;

[0010] (2) When using the extended sensor data method, the longitudinal acceleration output signal is used as the measurement feedback; the driving wheel rotation angular velocity signal and the wheel rolling radius are combined to predict the original wheel speed; the lateral acceleration signal is integrated to obtain the vehicle lateral movement speed;

[0011] (3) The Kalman filter is constructed by using an improved cubature Kalman filter; wherein the improved cubature Kalman filter is based on the characteristics of rapid update of filter variance and rapid convergence of filter gain in the actual Kalman filter iteration process, improves the cubature Kalman filter algorithm, introduces a balanced weighting factor for numerical fusion, and reduces the estimation error caused by signal fluctuations; wherein the Kalman filter takes the longitudinal velocity of the vehicle as the state variable, the longitudinal acceleration of the vehicle, the wheel rolling radius, the four-wheel turning angular velocity and the yaw angular velocity as the measurement variables, and the front wheel angle as the control variable, and the state equation is as follows:

[0012]

[0013] Where v(k+1) represents the longitudinal speed state variable of the system at time k+1, v k represents the system state variable at time k, Г k represents the system transfer matrix, W(k) represents the system observation noise, m is the total mass of the car, v y is the lateral speed, γ is the yaw rate, δ is the front wheel turning angle, I z is the moment of inertia of the vehicle body around the Z axis, a is the distance from the center of mass to the front axle, b is the distance from the center of mass to the rear axle, B f is the front axle track, B r is the rear axle track, F x1 Indicates the longitudinal force of the left front wheel, F x2 represents the longitudinal force of the right front wheel, F x3 Indicates the longitudinal force of the left rear wheel, F x4 Indicates the longitudinal force of the right rear wheel, F y1 Indicates the lateral force of the left front wheel, F y2 Indicates the lateral force of the right front wheel, F y3 Indicates the lateral force of the left rear wheel, F y4 Indicates the lateral force on the right rear wheel.

[0014] (4) According to the state equation described in step 1 (3), time updating is performed according to the improved volumetric Kalman filter principle. The time updating includes volume point calculation, state prediction and volume point updating;

[0015] The volume point calculation process is as follows:

[0016] In the Cartesian coordinate system, the integral expression of the spherical radial integral is:

[0017]

[0018] Where Un is the surface of the sphere, r is the radius of the sphere, exp(-r 2 ) is the weight equation, σ is the integration unit;

[0019] Using the Gass-Hermite criterion and the third-order spherical-radial volume criterion, the volume point integral is obtained as:

[0020]

[0021] In the formula, R n is the integration area, n is the vector dimension, ξ i is a volume point set.

[0022] The state prediction process is as follows:

[0023] State variable prediction:

[0024]

[0025] In the formula, Г k is the state transfer matrix;

[0026] The forecast error covariance is:

[0027]

[0028] Where W q is the external interference noise, which is a zero-mean, independent Gaussian distribution random signal.

[0029] The volume point update process is as follows:

[0030] According to the statistical characteristics of the state variables, the new volume point is:

[0031]

[0032] In the formula, x i k+1 / k represents the i-th new volume point, S k+1 / k The standard deviation of volume point update time.

[0033] (5) According to the state equation described in step 1 (3), measurement update is performed according to the improved cubature Kalman filter principle. The measurement update includes measurement variable update, measurement error update and filter gain update;

[0034] Among them, the measurement variable update process is as follows:

[0035] Substitute the new volume point into the measurement equation for nonlinear function transfer, and the measurement variable is:

[0036]

[0037] In the formula, is the updated measurement variable.

[0038] The measurement error update process is as follows:

[0039] The updated measurement error covariance is:

[0040]

[0041] Where W r is the covariance matrix, P γ k+1 is the measurement error covariance;

[0042] The cross-covariance of the updated state variables and measurement variables is:

[0043]

[0044] Where P xγ k+1 is the updated mutual covariance;

[0045] Among them, the updated filter gain is:

[0046]

[0047] In the formula, K k+1 Update the filter gain for the measurement.

[0048] (6) Based on the measurement update described in step 1 (5), the cubature Kalman filter algorithm is improved based on the characteristic of rapid convergence of the filter gain in the actual Kalman filter iteration process, and a balanced weighting factor is introduced for numerical fusion to reduce the error caused by signal fluctuations. The longitudinal vehicle speed observation value is obtained by combining the iterative filter value:

[0049]

[0050] In the formula, α is the balance weighting factor, is the longitudinal vehicle speed filter value, is the observed longitudinal vehicle speed.

[0051] Step 2: First, by analyzing the driving wheel speed signal described in step 1 (1), different fuzzy logic rules are formulated according to different working conditions (emergency driving, emergency braking, and coasting with gears), and then the similarity coefficients between the four driving wheels and the corresponding working conditions are calculated, and the reference speed is adaptively selected. Finally, the estimated value of the longitudinal vehicle speed is obtained by fusion integration:

[0052] (1) First, fuzzy logic inference rules are formulated for the complex nonlinear vehicle operating conditions of emergency driving, emergency braking, and coasting with gear. The fuzzy logic inference rules use x to represent the similarity between the current operating condition and the three complex nonlinear vehicle operating conditions of emergency driving, emergency braking, and coasting with gear. k Indicates that x k The fuzzy domain is [CS SS VS], corresponding to general similarity, similarity, and very similarity, with values ​​of 0.3, 0.8, and 1, respectively, where k = 1, 2, and 3 represent emergency driving conditions, emergency braking conditions, and coasting conditions with gear respectively;

[0053] (2) Performing working condition identification and judgment, taking the emergency braking and deceleration working condition as an example for analysis, when the vehicle is in the emergency braking working condition, all four driving wheels may slip or lock significantly. According to the change law of the wheel speed under this working condition, the corresponding characteristics of the current vehicle acceleration and the working condition are characterized in a similarity manner while ensuring stability, and the four driving wheels are respectively put into the judgment state, and the corresponding similarity coefficients are output;

[0054] (3) If the current wheel is already in the judgment state or has been in the judgment state for longer than the time threshold, the current wheel exits the judgment state and the next priority wheel enters the judgment state;

[0055] (4) If none of the four driving wheels enters the judgment state, it indicates that the current vehicle operating condition is normal and the judgment state ends;

[0056] (5) Based on the similarity coefficients outputted after judgment of the four driving wheels, the weighted average method is used to obtain the reference wheel speed under the current working condition:

[0057]

[0058] In the formula, w opt is the reference speed under the current working condition, w optk is the speed under different working conditions, u k is the similarity coefficient.

[0059] (6) Based on the reference vehicle speed under the current working condition, the longitudinal vehicle speed estimate is obtained by integration fusion:

[0060]

[0061] Where R is the rolling radius of the wheel under the current working condition, V x ′ is the estimated longitudinal vehicle speed.

[0062] Step 3: Perform a comprehensive analysis based on the longitudinal speed observation and estimated value, and combine the operating condition factors to obtain the final longitudinal speed through weighted fusion:

[0063] (1) Based on the comprehensive analysis of the longitudinal vehicle speed observation value based on the improved cubature Kalman filter and the longitudinal vehicle speed estimation value based on the fuzzy logic working condition identification, a working condition factor is proposed; wherein the working condition factor is represented by ε, and its value is determined by the wheel slip state. If the wheel is completely locked, ε=0; if the wheel slips greatly, 0<ε<1; if the wheel does not slip, ε=1;

[0064] (2) According to the operating condition factors, the observed and estimated values ​​of the longitudinal speed are weighted and fused to obtain the final longitudinal speed estimation result:

[0065]

[0066] In the formula, is the longitudinal speed observation value, V x ′ is the estimated longitudinal vehicle speed, and ε is the operating condition factor.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] 1. The present invention provides a method for estimating the longitudinal vehicle speed of a four-wheel drive hub motor vehicle, proposes an improved cubature Kalman filter method, adds a balance weighting factor, and comprehensively considers the influence of the fluctuation of the rotational angular velocity, longitudinal acceleration, lateral acceleration, yaw angular velocity and front wheel angle signal during the actual test process on the longitudinal vehicle speed estimation. Therefore, under complex working conditions, especially in complex nonlinear strong coupling systems, the method has strong adaptability and can stably and accurately realize the estimation of the longitudinal vehicle speed.

[0069] 2. The present invention provides a method for estimating the longitudinal speed of a four-wheel drive hub motor vehicle and a driving condition judgment method, which, after a complete optimization setting and fuzzy logic reasoning process, ensures the accuracy of driving condition identification and improves the accuracy of longitudinal speed estimation.

[0070] 3. The present invention provides a method for estimating the longitudinal speed of a four-wheel drive hub motor vehicle. By introducing operating condition factors, the method comprehensively considers the impact of different operating conditions on the vehicle state, especially for operating conditions that are easily affected by external interference and cannot accurately obtain the longitudinal speed. The method has good robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The present invention will be further described below in conjunction with the accompanying drawings:

[0072] Figure 1 Flowchart of the overall scheme for longitudinal vehicle speed estimation

[0073] Figure 2 Flowchart of longitudinal vehicle speed observation value calculation

[0074] Figure 3 Improved cubature Kalman filter flow chart

[0075] Figure 4 Longitudinal vehicle speed estimation calculation flow chart

[0076] Figure 5 Vehicle driving condition judgment flow chart

[0077] Figure 6 Flowchart of calculation of the final estimation result of longitudinal vehicle speed DETAILED DESCRIPTION

[0078] refer to Figure 1 , the present invention is implemented by adopting the following technical solutions:

[0079] Step 1: Reference Figure 2 First, the vehicle state parameters are collected, and the extended sensor data method is used. Then the cubature Kalman filter method is improved, and the balance weighting factor is added. Finally, the longitudinal speed observation value is obtained through weighted fusion:

[0080] (1) The vehicle mass, body moment of inertia, vehicle wheelbase and wheel track configuration parameters are measured by on-board sensors, and the speed, angular velocity, longitudinal acceleration, lateral acceleration, yaw angular velocity and front wheel angle state parameters of the four driving wheels are collected through the vehicle CAN bus and FlexRay bus;

[0081] (2) When using the extended sensor data method, the longitudinal acceleration output signal is used as the measurement feedback; the driving wheel rotation angular velocity signal and the wheel rolling radius are combined to predict the original wheel speed; the lateral acceleration signal is integrated to obtain the vehicle lateral movement speed;

[0082] (3) The Kalman filter is constructed by using an improved cubature Kalman filter; wherein the improved cubature Kalman filter is based on the characteristics of rapid update of filter variance and rapid convergence of filter gain in the actual Kalman filter iteration process, improves the cubature Kalman filter algorithm, introduces a balanced weighting factor for numerical fusion, and reduces the estimation error caused by signal fluctuations; wherein the Kalman filter takes the longitudinal velocity of the vehicle as the state variable, the longitudinal acceleration of the vehicle, the wheel rolling radius, the four-wheel turning angular velocity and the yaw angular velocity as the measurement variables, and the front wheel angle as the control variable, and the state equation is as follows:

[0083]

[0084] Where v(k+1) represents the longitudinal speed state variable of the system at time k+1, v k represents the system state variable at time k, Г k represents the system transfer matrix, W(k) represents the system observation noise, m is the total mass of the car, v yis the lateral speed, γ is the yaw rate, δ is the front wheel turning angle, I z is the moment of inertia of the vehicle body around the Z axis, a is the distance from the center of mass to the front axle, b is the distance from the center of mass to the rear axle, B f is the front axle track, B r is the rear axle track, F x1 Indicates the longitudinal force of the left front wheel, F x2 represents the longitudinal force of the right front wheel, F x3 Indicates the longitudinal force of the left rear wheel, F x4 Indicates the longitudinal force of the right rear wheel, F y1 Indicates the lateral force of the left front wheel, F y2 Indicates the lateral force of the right front wheel, F y3 Indicates the lateral force of the left rear wheel, F y4 Indicates the lateral force on the right rear wheel.

[0085] (4)Reference Figure 3 According to the state equation described in step 1 (3), time update is performed according to the improved volumetric Kalman filter principle. The time update includes volume point calculation, state prediction and volume point update;

[0086] The volume point calculation process is as follows:

[0087] In the Cartesian coordinate system, the integral expression of the spherical radial integral is:

[0088]

[0089] Where Un is the surface of the sphere, r is the radius of the sphere, exp(-r 2 ) is the weight equation, σ is the integration unit;

[0090] Using the Gass-Hermite criterion and the third-order spherical-radial volume criterion, the volume point integral is obtained as:

[0091]

[0092] In the formula, R n is the integration area, n is the vector dimension, ξ i is a volume point set.

[0093] The state prediction process is as follows:

[0094] State variable prediction:

[0095]

[0096] In the formula, Г k is the state transfer matrix;

[0097] The forecast error covariance is:

[0098]

[0099] Where W q is the external interference noise, which is a zero-mean, independent Gaussian distribution random signal.

[0100] The volume point update process is as follows:

[0101] According to the statistical characteristics of the state variables, the new volume point is:

[0102]

[0103] In the formula, x i k+1 / k represents the i-th new volume point, S k+1 / k The standard deviation of volume point update time.

[0104] (5)Reference Figure 3 According to the state equation described in step 1 (3), measurement update is performed according to the improved cubature Kalman filter principle. The measurement update includes measurement variable update, measurement error update and filter gain update;

[0105] Among them, the measurement variable update process is as follows:

[0106] Substitute the new volume point into the measurement equation for nonlinear function transfer, and the measurement variable is:

[0107]

[0108] In the formula, is the updated measurement variable.

[0109] The measurement error update process is as follows:

[0110] The updated measurement error covariance is:

[0111]

[0112] Where W r is the covariance matrix, P γ k+1 is the measurement error covariance;

[0113] The cross-covariance of the updated state variables and measurement variables is:

[0114]

[0115] Where P xγ k+1 is the updated mutual covariance;

[0116] Among them, the updated filter gain is:

[0117]

[0118] In the formula, K k+1 Update the filter gain for the measurement.

[0119] (6) Based on the measurement update described in step 1 (5), the cubature Kalman filter algorithm is improved based on the characteristic of rapid convergence of the filter gain in the actual Kalman filter iteration process, and a balanced weighting factor is introduced for numerical fusion to reduce the error caused by signal fluctuations. The longitudinal vehicle speed observation value is obtained by combining the iterative filter value:

[0120]

[0121] In the formula, α is the balance weighting factor, is the longitudinal vehicle speed filter value, is the observed longitudinal vehicle speed.

[0122] Step 2: Reference Figure 4 First, by analyzing the driving wheel speed signal described in step 1 (1), different fuzzy logic rules are formulated according to different working conditions (emergency driving, emergency braking, and coasting with gears), and then the similarity coefficients between the four driving wheels and the corresponding working conditions are calculated, and the reference speed is adaptively selected. Finally, the estimated value of the longitudinal vehicle speed is obtained by fusion integration:

[0123] (1) First, fuzzy logic inference rules are formulated for the complex nonlinear vehicle operating conditions of emergency driving, emergency braking, and coasting with gear. The fuzzy logic inference rules use x to represent the similarity between the current operating condition and the three complex nonlinear vehicle operating conditions of emergency driving, emergency braking, and coasting with gear. k Indicates that x k The fuzzy domain is [CS SS VS], corresponding to general similarity, similarity, and very similarity, with values ​​of 0.3, 0.8, and 1, respectively, where k = 1, 2, and 3 represent emergency driving conditions, emergency braking conditions, and coasting conditions with gear respectively;

[0124] (2)Reference Figure 5 , perform working condition identification and judgment, take the emergency braking deceleration working condition as an example for analysis. When the vehicle is in an emergency braking working condition, the four driving wheels may slip or lock greatly. According to the change law of the wheel speed under this working condition, the corresponding characteristics of the current vehicle acceleration and the working condition are characterized in a similar way under the premise of ensuring stability, and the four driving wheels are respectively put into the judgment state, and the corresponding similarity coefficient is output;

[0125] (3) If the current wheel is already in the judgment state or has been in the judgment state for longer than the time threshold, the current wheel exits the judgment state and the next priority wheel enters the judgment state;

[0126] (4) If none of the four driving wheels enters the judgment state, it indicates that the current vehicle operating condition is normal and the judgment state ends;

[0127] (5) Based on the similarity coefficients outputted after judgment of the four driving wheels, the weighted average method is used to obtain the reference wheel speed under the current working condition:

[0128]

[0129] In the formula, w opt is the reference speed under the current working condition, w optk is the speed under different working conditions, u k is the similarity coefficient.

[0130] (6) Based on the reference vehicle speed under the current working condition, the longitudinal vehicle speed estimate is obtained by integration fusion:

[0131]

[0132] Where R is the rolling radius of the wheel under the current working condition, V x ′ is the estimated longitudinal vehicle speed.

[0133] Step 3: Reference Figure 6 , based on the observed and estimated longitudinal vehicle speeds, a comprehensive analysis is performed, combined with the operating condition factors, and the final longitudinal vehicle speed is obtained through weighted fusion:

[0134] (1) Based on the comprehensive analysis of the longitudinal vehicle speed observation value based on the improved cubature Kalman filter and the longitudinal vehicle speed estimation value based on the fuzzy logic working condition identification, a working condition factor is proposed; wherein, the working condition factor is represented by ε, and its value is determined by the wheel slip state. If the wheel is completely locked, ε = 0; if the wheel slips greatly, 0 < ε < 1; if the wheel does not slip, ε = 1; taking the vehicle entering the emergency braking condition as an example for analysis, at this time, the wheel slips greatly, and the value of the working condition factor ε is 0 < ε < 1;

[0135] (2) According to the operating condition factors, the observed and estimated values ​​of the longitudinal speed are weighted and fused to obtain the final longitudinal speed estimation result:

[0136]

[0137] In the formula, is the longitudinal speed observation value, V x ′ is the estimated longitudinal vehicle speed, and ε is the operating condition factor.

Claims

1. A method for estimating the longitudinal speed of a four-wheel drive hub motor vehicle, characterized in that: The method includes the following steps: Step 1: First, collect vehicle state parameters, use the extended sensor data method, then improve the volumetric Kalman filter method, add a balance weighting factor, and finally obtain the longitudinal speed observation value through weighted fusion: (1) obtaining vehicle configuration parameters according to vehicle-mounted sensors and collecting vehicle status parameters through a data bus; wherein the vehicle configuration parameters include vehicle mass, vehicle body moment of inertia, vehicle wheelbase and wheel track; the vehicle status parameters include four driving wheel speeds, driving wheel rotation angular velocity, longitudinal acceleration, lateral acceleration, yaw angular velocity and front wheel steering angle; the data bus includes a vehicle CAN bus and a FlexRay bus; (2) When using the extended sensor data method, the longitudinal acceleration output signal is used as the measurement feedback; the driving wheel rotation angular velocity signal and the wheel rolling radius are combined to predict the original wheel speed; the lateral acceleration signal is integrated to obtain the vehicle lateral movement speed; (3) The Kalman filter is constructed by using an improved cubature Kalman filter; wherein the improved cubature Kalman filter is based on the characteristics of rapid update of filter variance and rapid convergence of filter gain in the actual Kalman filter iteration process, improves the cubature Kalman filter algorithm, introduces a balanced weighting factor for numerical fusion, and reduces the estimation error caused by signal fluctuations; wherein the Kalman filter takes the longitudinal velocity of the vehicle as the state variable, the longitudinal acceleration of the vehicle, the wheel rolling radius, the four-wheel turning angular velocity and the yaw angular velocity as the measurement variables, and the front wheel angle as the control variable, and the state equation is as follows: Where v(k+1) represents the longitudinal speed state variable of the system at time k+1, v k represents the system state variable at time k, Г k represents the system transfer matrix, W(k) represents the system observation noise, m is the total mass of the car, v y is the lateral speed, γ is the yaw rate, δ is the front wheel turning angle, I z is the moment of inertia of the vehicle body around the Z axis, a is the distance from the center of mass to the front axle, b is the distance from the center of mass to the rear axle, B f is the front axle track, B r is the rear axle track, F x1 Indicates the longitudinal force of the left front wheel, F x2 represents the longitudinal force of the right front wheel, F x3 Indicates the longitudinal force of the left rear wheel, F x4 Indicates the longitudinal force of the right rear wheel, F y1 Indicates the lateral force of the left front wheel, F y2 Indicates the lateral force of the right front wheel, F y3 Indicates the lateral force of the left rear wheel, F y4 Indicates the lateral force of the right rear wheel; (4) performing time updating according to the state equation described in step 1 (3) and the improved volumetric Kalman filter principle; wherein the time updating includes volume point calculation, state prediction and volume point updating; The volume point calculation process is as follows: In the Cartesian coordinate system, the integral expression of the spherical radial integral is: Where Un is the surface of the sphere, r is the radius of the sphere, exp(-r 2 ) is the weight equation, σ is the integration unit; Using the Gass-Hermite criterion and the third-order spherical-radial volume criterion, the volume point integral is obtained as: In the formula, R n is the integration area, n is the vector dimension, ξ i is the volume point set; The state prediction process is as follows: State variable prediction: In the formula, Г k is the state transfer matrix; The forecast error covariance is: Where W q is the external interference noise, which is a zero-mean, independent Gaussian distribution random signal; The volume point update process is as follows: According to the statistical characteristics of the state variables, the new volume point is: In the formula, x i k+1 / k represents the i-th new volume point, S k+1 / k is the standard deviation of the volume point update; (5) performing measurement update according to the state equation described in step 1 (3) and the improved cubature Kalman filter principle; wherein the measurement update includes measurement variable update, measurement error update and filter gain update; Among them, the measurement variable update process is as follows: Substitute the new volume point into the measurement equation for nonlinear function transfer, and the measurement variable is: In the formula, is the updated measurement variable; The measurement error update process is as follows: The updated measurement error covariance is: Where W r is the covariance matrix, P γ k+1 is the measurement error covariance; The cross-covariance of the updated state variables and measurement variables is: Where P xγ k+1 is the updated mutual covariance; Among them, the updated filter gain is: In the formula, K k+1 Update the filter gain for the measurement; (6) Based on the measurement update described in step 1 (5), the cubature Kalman filter algorithm is improved based on the characteristic of rapid convergence of the filter gain in the actual Kalman filter iteration process, and a balanced weighting factor is introduced for numerical fusion to reduce the error caused by signal fluctuations. The longitudinal vehicle speed observation value is obtained by combining the iterative filter value: In the formula, α is the balance weighting factor, is the longitudinal vehicle speed filter value, is the observed longitudinal vehicle speed; Step 2: First, by analyzing the driving wheel speed signal described in step 1 (1), different fuzzy logic rules are formulated according to different working conditions (emergency driving, emergency braking, and coasting with gears), and then the similarity coefficients between the four driving wheels and the corresponding working conditions are calculated, and the reference speed is adaptively selected. Finally, the estimated value of the longitudinal vehicle speed is obtained by fusion integration: (1) First, fuzzy logic inference rules are formulated for the complex nonlinear vehicle operating conditions of emergency driving, emergency braking, and coasting with gear. The fuzzy logic inference rules use x to represent the similarity between the current operating condition and the three complex nonlinear vehicle operating conditions of emergency driving, emergency braking, and coasting with gear. k Indicates that x k The fuzzy domain is [CS SS VS], corresponding to general similarity, similarity, and very similarity, with values ​​of 0.3, 0.8, and 1, respectively, where k = 1, 2, and 3 represent emergency driving conditions, emergency braking conditions, and coasting conditions with gear respectively; (2) Performing working condition identification and judgment, taking the emergency braking and deceleration working condition as an example for analysis, when the vehicle is in the emergency braking working condition, all four driving wheels may slip or lock significantly. According to the change law of the wheel speed under this working condition, the corresponding characteristics of the current vehicle acceleration and the working condition are characterized in a similarity manner while ensuring stability, and the four driving wheels are respectively put into the judgment state, and the corresponding similarity coefficients are output; (3) If the current wheel is already in the judgment state or has been in the judgment state for longer than the time threshold, the current wheel exits the judgment state and the next priority wheel enters the judgment state; (4) If none of the four driving wheels enters the judgment state, it indicates that the current vehicle operating condition is normal and the judgment state ends; (5) Based on the similarity coefficients outputted after judgment of the four driving wheels, the weighted average method is used to obtain the reference wheel speed under the current working condition: In the formula, w opt is the reference speed under the current working condition, w optk is the speed under different working conditions, u k is the similarity coefficient; (6) Based on the reference vehicle speed under the current working condition, the longitudinal vehicle speed estimate is obtained by integration fusion: Where R is the rolling radius of the wheel under the current working condition, V x ′ is the estimated longitudinal vehicle speed; Step 3: Comprehensively analyze the longitudinal vehicle speed observation value obtained in step 1 and the longitudinal vehicle speed estimation value obtained in step 2, and combine the working condition factors to obtain the final longitudinal vehicle speed through weighted fusion: (1) Based on the comprehensive analysis of the longitudinal vehicle speed observation value based on the improved cubature Kalman filter and the longitudinal vehicle speed estimation value based on the fuzzy logic working condition identification, a working condition factor is proposed; wherein, the working condition factor is represented by ε, and its value is determined by the wheel slip state. If the wheel is completely locked, ε = 0; if the wheel slips greatly, 0 < ε < 1; if the wheel does not slip, ε = 1; taking the vehicle entering the emergency braking condition as an example for analysis, at this time, the wheel slips greatly, and the value of the working condition factor ε is 0 < ε < 1; (2) According to the operating condition factors, the observed and estimated values ​​of the longitudinal speed are weighted and fused to obtain the final longitudinal speed estimation result: In the formula, is the longitudinal speed observation value, V x ′ is the estimated longitudinal vehicle speed, and ε is the operating condition factor.

Images