A method for estimating the slope of automobile roads based on information fusion

Abstract

The present invention aims to solve the problem that the existing automobile slope estimation method cannot take into account low cost, low calibration difficulty, high accuracy, and strong real-time performance at the same time, and proposes a method for estimating the slope of an automobile road based on information fusion. The method includes the following steps: the first step is to obtain the corresponding signal from the bus to ensure that the available information is fully utilized to improve the accuracy; the second step is to perform low-pass filtering and other processing on the original signal, and at the same time solve the signal needed later; the third step is to establish the longitudinal dynamic model of the whole vehicle and the longitudinal kinematic model of the road slope according to the parameters of the first step and the second step; the fourth step is to use the nonlinear observer and the least square method with forgetting factor to solve the slope in the dynamic and kinematic conditions respectively; the fifth step is to fuse the two methods to obtain the final slope estimation result.

Description

Technical Field

[0001] The invention relates to a road slope recognition method for automobiles, and in particular to a road slope real-time estimation method integrating kinematics and dynamics methods. Background Art

[0002] As one of the important components of vehicle driving state parameters, slope is very important for vehicle stability and economic control. In actual driving, road slope directly affects the vehicle's driving state and the driver's behavior, and the slope estimation result with high accuracy can provide guarantee for vehicle control.

[0003] At present, the research on slope recognition mainly uses sensors to directly collect relevant data, and then obtains the estimated result of the road slope through mathematical algorithms. However, in practical applications, the accuracy of sensors is easily affected by the environment and errors occur, which in turn lead to incorrect estimation results. At the same time, some high-precision sensors are expensive and therefore not suitable for mass production. Therefore, estimating the slope of the car using the sensor information of the vehicle's standard configuration is the main problem at present.

[0004] In recent years, in order to meet the actual control needs of vehicles, domestic scholars have conducted a series of studies on vehicle slope estimation. For example, the Chinese patent publication number is CN113978473A, and the publication date is 2022-01-28, which discloses a vehicle mass and road slope estimation method. This method integrates the longitudinal dynamic model of the vehicle mass and road slope and the longitudinal kinematic model of the road slope. When predicting the estimated value of the vehicle mass and the estimated value of the road slope, it can simultaneously estimate the two situations. Chinese patent publication number is CN111391857B, and the publication date is 2021-08-06, which discloses a distributed drive electric vehicle slope estimation method and system. The system uses a Kalman filter algorithm to filter the longitudinal acceleration of the vehicle, and calculates the slope of the road surface on which the vehicle is located based on the longitudinal acceleration of the vehicle obtained by filtering. However, among the mainstream methods of existing patents, the complex method using multiple sensors has a high installation cost, a large amount of calibration, and is difficult to apply in practice; the method using conventional sensors tends to ignore the original signal processing part, and does not make full use of the signals on the vehicle bus, resulting in low estimation accuracy. In view of the current problems in vehicle slope estimation, it is of great significance to develop a corresponding slope estimation method with high real-time performance, high precision, low cost and good adaptability to working conditions, which is of great significance to vehicle stability and economic control. Summary of the invention

[0005] The present invention aims to solve the problem that the existing automobile slope estimation method cannot take into account low cost, low calibration difficulty, high accuracy and strong real-time performance at the same time, and proposes an automobile road slope estimation method based on information fusion.

[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solution: comprising the following steps:

[0007] The first step is to obtain the necessary signals from the vehicle bus

[0008] First, determine the signal required for slope estimation: the acceleration ax of the vehicle in the x, y, and z directions obtained by the inertial sensor sensor ,ay sensor ,az sensor ; Vehicle speed v; Vehicle current gear information; Engine output torque T tq ; The total mass of the vehicle at this time is m; The total braking force of the vehicle wheels is F brk .

[0009] The second step is to process the signal obtained in the first step.

[0010] The original signals collected by the sensors contain incorrect information such as sharp points and burrs due to vehicle gear shifting, acceleration and deceleration, so they need to be preliminarily processed to remove some distorted and erroneous information.

[0011] The signal from the bus is used to calculate and obtain the required data that cannot be directly observed, including: gravitational acceleration g; vehicle gearbox transmission ratio i g ; Transmission system transmission efficiency η; Rolling resistance coefficient f; Vehicle acceleration a.

[0012] Among them, the value of gravity acceleration g is related to the geographical location during vehicle driving, so gravity acceleration g needs to be recalculated before use. The specific calculation formula is:

[0013]

[0014] Vehicle gearbox transmission ratio i g The transmission efficiency η of the transmission system is solved by looking up the table, the rolling resistance coefficient f is calculated using the empirical formula of the tire rolling resistance coefficient; the vehicle acceleration a is calculated by derivation of the vehicle speed v.

[0015] The third step is to establish the vehicle longitudinal dynamics model and longitudinal kinematics model

[0016] According to the parameters of the first and second steps, the longitudinal dynamic model of the whole vehicle and the longitudinal kinematic model about the road slope are established.

[0017] The longitudinal dynamics model is as follows:

[0018]

[0019] Where, r is the wheel radius

[0020] α——The slope of the road where the vehicle is at this time

[0021] C D ——Air resistance coefficient

[0022] A——Vehicle frontal area

[0023] δ——Vehicle rotation mass conversion factor

[0024] Among them, the longitudinal kinematic model is specifically:

[0025] ax sensor =a+gsinα (3)

[0026] The fourth step is to estimate the road slope based on the longitudinal dynamic model and the longitudinal kinematic model respectively.

[0027] According to the longitudinal dynamic model in the third step, the nonlinear observer is used to solve and obtain the slope estimation value α1 based on dynamics; according to the longitudinal kinematic model in the third step, the recursive least square method with forgetting factor is used to solve and obtain the slope estimation value α2 based on kinematics.

[0028] Step 5: Fusion of two types of slope information

[0029] Set the weights that change with speed, fuse the two types of slope information α1 and α2, and get the final road slope estimate Output as the final slope estimate.

[0030] In the second step of the present invention, processing the acquired signal refers to:

[0031] For raw sensor signals with many sharp points and instability, the following processing is required: low-pass filtering of raw signals; signal value limitation; signal change rate limitation.

[0032] The low-pass filter formula is:

[0033] x(k)=(1-alpha)·a(k)+alpha·x(k-1) (4)

[0034] Where x(k) is the output of the kth low-pass filter.

[0035] alpha——low-pass filter coefficient

[0036] The numerical limit formula is:

[0037]

[0038] In the formula, x is the signal to be processed

[0039] x max——Numerical limit maximum value

[0040] x min ——Minimum value of numerical limit

[0041] The rate-of-change limiting formula is:

[0042]

[0043] In the formula, k is the rate of change of the signal to be processed

[0044] k max ——Change rate limit maximum value

[0045] k min ——Minimum value of rate of change limit

[0046] In the fourth step of the present invention, solving the nonlinear observer means:

[0047] Considering the nonlinearity of the vehicle longitudinal dynamics equation, a nonlinear observer is selected to observe the slope value; according to the vehicle longitudinal dynamics model, the difference between the measured value of the vehicle longitudinal velocity and the estimated value of the longitudinal velocity is selected as the correction term, and its nonlinear full-dimensional state observer is listed as:

[0048]

[0049]

[0050] In the formula, ——Estimated vehicle speed

[0051] K——Observer gain coefficient

[0052] In the fourth step of the present invention, the recursive least square method with forgetting factor refers to:

[0053] Transform formula (3) into the following form:

[0054] gsinα=ax sensor -a (8)

[0055] Therefore, the slope estimation value can be solved as a system identification problem, where the slope sine value sinα is used as the system parameter to be identified θ, and the g term is used as the observation input ax sensor -a item as output y.

[0056] The slope is calculated using the recursive least squares method with a forgetting factor, and the recursive formula is used to solve it:

[0057]

[0058] In the formula, K n ——Gain coefficient at time n

[0059] P n ——N-time covariance

[0060] λ——Forgetting factor

[0061] Finally, the kinematically based slope estimation value α2 is obtained;

[0062] In the fifth step of the present invention, information fusion refers to:

[0063] The proportional coefficient is set according to the vehicle speed to complete the information fusion between dynamics and kinematics. The formula is as follows:

[0064]

[0065] In the formula, k1 is the fusion weight coefficient of α1

[0066] k2——fusion weight coefficient of α2

[0067] The specific solution process of k1 and k2 is as follows:

[0068]

[0069] k2=1-k1 (10)

[0070] In the formula, a, b——k1 polynomial coefficients, it is necessary to ensure When k1 = m

[0071] v re f——Reference threshold speed, which can be calibrated according to the differences between actual vehicles

[0072] m——The maximum value of the k1 value range, which can be calibrated according to the differences between actual vehicles

[0073] Through this method, it can be ensured that the proportion of the two estimation methods changes with the vehicle speed, and the final result is more reliable.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] 1. The method for estimating the slope of a vehicle road based on information fusion described in the present invention is an estimation method based on the fusion of kinematics and dynamics multi-information, which does not require the use of expensive high-precision sensors and can ensure the real-time and accuracy requirements of slope estimation in actual work;

[0076] 2. The method for estimating the slope of a road based on information fusion described in the present invention can adjust the weights of the dynamic estimation result and the kinematic estimation result according to the vehicle speed, thereby improving the adaptability of the slope estimation method to working conditions. Therefore, the slope information based on the fusion of the two types of information is more reliable.

[0077] 3. The method for estimating the slope of an automobile road based on information fusion described in the present invention uses a nonlinear observer and a least square method with a forgetting factor to estimate the slope based on dynamics and kinematics respectively. Compared with the traditional method, the calibration amount is less and it can meet the real-time requirements, which is conducive to practical application;

[0078] 4. The automobile road slope estimation method based on information fusion described in the present invention processes the original signal of the sensor, effectively reducing the influence of gear shifting and rapid acceleration and deceleration on the result accuracy;

[0079] 5. The automobile road slope estimation method based on information fusion described in the present invention makes full use of bus signals, thereby ensuring the rapidity of the slope estimation method and improving the accuracy of the estimation result. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The present invention will be further described below in conjunction with the accompanying drawings:

[0081] Figure 1 It is a logic diagram of the road slope estimation method based on road information fusion according to the present invention;

[0082] Figure 2 A schematic diagram of a method for solving a dynamic slope estimation value by a nonlinear state observer according to the present invention;

[0083] Figure 3 The figure is a schematic diagram of solving the recursive least square method with forgetting factor described in the present invention.

[0084] Figure 4 It is a schematic diagram of the weight coefficient of the present invention changing with speed. DETAILED DESCRIPTION

[0085] The present invention is described in detail below in conjunction with the accompanying drawings:

[0086] The invention discloses a method for estimating the slope of a vehicle road based on information fusion. By obtaining the corresponding available signal from the bus and filtering the original signal, the slopes of the vehicle longitudinal dynamics model and the vehicle longitudinal kinematics model are solved respectively according to the nonlinear observer and the least square method with a forgetting factor, and then the two methods are fused according to the vehicle speed information to obtain the final slope estimation result. The basic principle of the method is to reduce the dependence of the estimation model on the modeling accuracy by considering the model information in the historical data, and at the same time adopt the longitudinal dynamics and longitudinal kinematics two ways, and adjust the weight between the two in real time through the vehicle speed to enhance the working condition applicability of the method. Its core includes a nonlinear state observer and a recursive least square method with a forgetting factor. The method can estimate the slope of the vehicle under the conditions of low cost, low calibration difficulty, high accuracy, and strong real-time performance, and effectively improve the stability and economy in the vehicle control. Next, the design and estimation ideas of the method for estimating the slope of a vehicle road based on information fusion are specifically described.

[0087] See also Figure 1 The method for estimating the slope of a road based on information fusion described in the present invention includes five parts: signal acquisition, signal processing, model building, specific solution and information fusion. The method for estimating the slope of a road based on information fusion is described in detail step by step below.

[0088] The following steps are involved:

[0089] The first step is to obtain the necessary signals from the vehicle bus ( Figure 1 1. Shown in the position)

[0090] First, determine the signal required for slope estimation: the acceleration ax around the vehicle in the x, y, and z directions obtained by the inertial sensor sensor ,ay sensor ,az sensor ; Vehicle speed v; Vehicle current gear information; Engine output torque T tq ; The total mass of the vehicle at this time is m; The total braking force of the vehicle wheels is F brk .

[0091] The second step is to process the signal obtained in the first step ( Figure 1 2. Shown in the position)

[0092] The original signals collected by the sensors contain incorrect information such as sharp points and burrs due to vehicle gear shifting, acceleration and deceleration, so they need to be preliminarily processed to remove some distorted and erroneous information.

[0093] The signal from the bus is used to calculate and obtain the required data that cannot be directly observed, including: gravitational acceleration g; vehicle gearbox transmission ratio i g; Transmission system transmission efficiency η; Rolling resistance coefficient f; Vehicle acceleration a.

[0094] Among them, the value of gravity acceleration g is related to the geographical location during vehicle driving, so gravity acceleration g needs to be recalculated before use. The specific calculation formula is:

[0095]

[0096] Vehicle gearbox transmission ratio i g The transmission efficiency η of the transmission system is solved by looking up the table, the rolling resistance coefficient f is calculated using the empirical formula of the tire rolling resistance coefficient; the vehicle acceleration a is calculated by derivation of the vehicle speed v.

[0097] The third step is to establish the vehicle longitudinal dynamics model and longitudinal kinematics model ( Figure 1 3. Shown in the position)

[0098] According to the parameters of the first and second steps, the longitudinal dynamic model of the whole vehicle and the longitudinal kinematic model about the road slope are established.

[0099] The longitudinal dynamics model is as follows:

[0100]

[0101] Where, r is the wheel radius

[0102] α——The slope of the road where the vehicle is at this time

[0103] C D ——Air resistance coefficient

[0104] A——Vehicle frontal area

[0105] δ——Vehicle rotation mass conversion factor

[0106] Among them, the longitudinal kinematic model is specifically:

[0107] ax sensor =a+gsinα (3)

[0108] The fourth step is to estimate the road slope based on the longitudinal dynamic model and the longitudinal kinematic model respectively ( Figure 1 4. Shown in the position)

[0109] According to the longitudinal dynamic model in the third step, the nonlinear observer is used to solve and obtain the slope estimation value α1 based on dynamics; according to the longitudinal kinematic model in the third step, the recursive least square method with forgetting factor is used to solve and obtain the slope estimation value α2 based on kinematics.

[0110] The fifth step is to integrate the two slope information ( Figure 1 5. Shown in the position)

[0111] Set the weights that change with speed, fuse the two types of slope information α1 and α2, and get the final road slope estimate Output as the final slope estimate.

[0112] In the second step of the present invention, processing the acquired signal refers to:

[0113] For raw sensor signals with many sharp points and instability, the following processing is required: low-pass filtering of raw signals; signal value limitation; signal change rate limitation.

[0114] The low-pass filter formula is:

[0115] x(k)=(1-alpha)·a(k)+alpha·x(k-1) (4)

[0116] Where x(k) is the output of the kth low-pass filter.

[0117] alpha——low-pass filter coefficient

[0118] The numerical limit formula is:

[0119]

[0120] In the formula, x is the signal to be processed

[0121] x max ——Numerical limit maximum value

[0122] x min ——Minimum value of numerical limit

[0123] The rate-of-change limiting formula is:

[0124]

[0125] In the formula, k is the rate of change of the signal to be processed

[0126] k max ——Change rate limit maximum value

[0127] k min ——Minimum value of rate of change limit

[0128] See also Figure 2 , which is a schematic diagram of a nonlinear state observer. In the fourth step of the present invention, solving the nonlinear state observer means:

[0129] Considering the nonlinearity of the vehicle longitudinal dynamics equation, a nonlinear observer is selected to observe the slope value; according to the vehicle longitudinal dynamics model, the difference between the measured value of the vehicle longitudinal velocity and the estimated value of the longitudinal velocity is selected as the correction term, and its nonlinear full-dimensional state observer is listed as:

[0130]

[0131]

[0132] In the formula, ——Estimated vehicle speed

[0133] K——Observer gain coefficient

[0134] See also Figure 3 , is the iterative process of recursive least squares method with forgetting factor. In the fourth step of the present invention, the recursive least squares method with forgetting factor refers to:

[0135] Transform formula (3) into the following form:

[0136] gsinα=ax sensor -a (8)

[0137] Therefore, the slope estimation value can be solved as a system identification problem, where the slope sine value sinα is used as the system parameter to be identified θ, and the g term is used as the observation input ax sensor -a item as output y.

[0138] The slope is calculated using the recursive least squares method with a forgetting factor, and the recursive formula is used to solve it:

[0139]

[0140] In the formula, K n ——Gain coefficient at time n

[0141] P n ——N-time covariance

[0142] λ——Forgetting factor

[0143] Finally, the kinematically based slope estimation value α2 is obtained;

[0144] In the fifth step of the present invention, information fusion refers to:

[0145] The proportional coefficient is set according to the vehicle speed to complete the information fusion between dynamics and kinematics. The formula is as follows:

[0146]

[0147] In the formula, k1 is the fusion weight coefficient of α1

[0148] k2——fusion weight coefficient of α2

[0149] The specific solution process of k1 and k2 is as follows:

[0150]

[0151] k2=1-k1 (10)

[0152] In the formula, a, b——k1 polynomial coefficients, it is necessary to ensure When k1 = m

[0153] v ref ——Reference threshold speed, which can be calibrated according to the differences between actual vehicles

[0154] m——The maximum value of the k1 value range, which can be calibrated according to the differences between actual vehicles

[0155] See also Figure 4 , is the relationship between the weight and the vehicle speed in the information fusion process. Through this method, it can be ensured that the proportion of the two estimation methods changes with the change of vehicle speed, and the final result is more reliable.

Claims

1. A method for estimating the slope of a road based on information fusion, characterized in that The following steps are involved: The first step is to obtain the necessary signals from the vehicle bus First, determine the signal required for slope estimation: the acceleration ax of the vehicle in the x, y, and z directions obtained by the inertial sensor sensor ,ay sensor ,az sensor ; Vehicle speed v; Vehicle current gear information; Engine output torque T tq ; The total mass of the vehicle at this time is m; Total braking force of vehicle wheels F brk ; The second step is to process the signal obtained in the first step. The original signals collected by the sensors contain incorrect information such as sharp points and burrs due to the vehicle's gear shifting, acceleration and deceleration, so they need to be processed initially to remove some distorted information and erroneous information. The signal from the bus is used to calculate and obtain the required data that cannot be directly observed, including: gravitational acceleration g; vehicle gearbox transmission ratio i g ; Transmission system transmission efficiency η; Rolling resistance coefficient f; Vehicle acceleration a; Among them, the value of gravity acceleration g is related to the geographical location during vehicle driving, so gravity acceleration g needs to be recalculated before use. The specific calculation formula is: Vehicle gearbox transmission ratio i g The transmission efficiency η of the transmission system is solved by looking up the table, and the rolling resistance coefficient f is calculated using the empirical formula of the tire rolling resistance coefficient; the vehicle acceleration a is calculated by taking the derivative of the vehicle speed v; The third step is to establish the vehicle longitudinal dynamics model and longitudinal kinematics model Establishing the longitudinal dynamics model of the vehicle and the longitudinal kinematics model about the road slope according to the parameters of the first and second steps; The longitudinal dynamics model is as follows: Where, r is the wheel radius α——The slope of the road where the vehicle is at this time C D ——Air resistance coefficient A——Vehicle frontal area δ——Vehicle rotation mass conversion factor Among them, the longitudinal kinematic model is specifically: ax sensor =a+gsinα (3) The fourth step is to estimate the road slope based on the longitudinal dynamic model and the longitudinal kinematic model respectively. According to the longitudinal dynamic model in the third step, the nonlinear observer is used to solve and obtain the slope estimation value α1 based on dynamics; according to the longitudinal kinematic model in the third step, the recursive least square method with forgetting factor is used to solve and obtain the slope estimation value α2 based on kinematics; Step 5: Fusion of two types of slope information Set the weights that change with speed, fuse the two types of slope information α1 and α2, and get the final road slope estimate Output as the final slope estimate.

2. The method for estimating the slope of a road based on information fusion according to claim 1, characterized in that: In the second step of the present invention, processing the acquired signal refers to: For raw sensor signals with many sharp points and instability, the following processing is required: low-pass filtering of raw signals; signal value limitation; signal change rate limitation; The low-pass filter formula is: x(k)=(1-alpha)·a(k)+alpha·x(k-1) (4) Where x(k) is the output of the kth low-pass filter. alpha——low-pass filter coefficient The numerical limit formula is: In the formula, x is the signal to be processed x max ——Numerical limit maximum value x min ——Minimum value of numerical limit The rate-of-change limiting formula is: In the formula, k is the rate of change of the signal to be processed k max ——Change rate limit maximum value k min ——Minimum value of rate of change limit After the above treatment, the second step of the method is carried out.

3. The method for estimating the slope of a road based on information fusion according to claim 1, characterized in that: In the fourth step of the present invention, solving the nonlinear observer means: Considering the nonlinearity of the vehicle longitudinal dynamics equation, a nonlinear observer is selected to observe the slope value; according to the vehicle longitudinal dynamics model, the difference between the measured value of the vehicle longitudinal velocity and the estimated value of the longitudinal velocity is selected as the correction term, and its nonlinear full-dimensional state observer is listed as: In the formula, ——Estimated vehicle speed K——Observer gain coefficient The nonlinear state observer is used to solve for α1 in the dynamic case.

4. The method for estimating the slope of a road based on information fusion according to claim 1, characterized in that: In the fourth step of the present invention, the recursive least square method with forgetting factor refers to: Transform formula (3) into the following form: gsinα=ax sensor -a (8) Therefore, the slope estimation value can be solved as a system identification problem, where the slope sine value sinα is used as the system parameter to be identified θ, and the g term is used as the observation input ax sensor -a item as output y; The slope is calculated using the recursive least squares method with a forgetting factor, and the recursive formula is used to solve it: In the formula, K n ——Gain coefficient at time n P n ——N-time covariance λ——Forgetting factor Finally, the kinematically based slope estimate α2 is obtained.

5. The method for estimating the slope of a road based on information fusion according to claim 1, characterized in that: In the fifth step of the present invention, information fusion refers to: The proportional coefficient is set according to the vehicle speed to complete the information fusion between dynamics and kinematics. The formula is as follows: In the formula, k1 is the fusion weight coefficient of α1 k2——fusion weight coefficient of α2 The specific solution process of k1 and k2 is as follows: k2=1-k1 (10) In the formula, a, b——k1 polynomial coefficients, it is necessary to ensure When k1 = m v ref ——Reference threshold speed, which can be calibrated according to the differences between actual vehicles m——The maximum value of the k1 value range, which can be calibrated according to the differences between actual vehicles Through this method, it can be ensured that the proportion of the two estimation methods changes with the vehicle speed, and the final result is more reliable.

Images