A wheel speed correction method for intelligent driving

Abstract

The application discloses a wheel speed correction method for intelligent driving, and the speed time coefficient calibration method under different road conditions and variable load conditions is used in an IMU+wheel speed fusion odometer scheme, wheel speed error is reduced, and the precision of the local odometer is improved. In addition, the final result is to generate a speed scale coefficient lookup table. Through calibration, the GNSS signal is fixed in the lookup table, the situation without GNSS can be met, and because there is no online calibration calculation, the computing power is saved, the computing power requirement of the operation platform is relatively low, and the embedded scheme is promoted.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent driving, and particularly relates to a wheel speed correction method for intelligent driving. BACKGROUND

[0002] In recent years, automatic driving technology has developed rapidly, and various sensors have been applied to cars. Some companies have also proposed positioning methods using inertial devices. Inertial devices can conveniently provide global positioning information of the current car. However, such inertial devices have a high degree of dependence on GNSS / GPS signals, and in some specific scenarios, GNSS / GPS signals cannot be well acquired. Therefore, there is a considerable challenge to positioning in these special scenarios. At present, positioning technology based on high-precision maps has been widely applied. As long as there is a high-precision map corresponding to the scene, a quite good positioning effect can be achieved. However, in some scenarios, the camera or laser radar sensor may not work reliably. Therefore, a stable working odometer is necessary. The sensor that can work stably and reliably must meet the condition of small dependence on the outside world. The sensor signals that can be considered include the signals of the car itself and the acceleration and angular velocity measured by the IMU (inertial test element).

[0003] In the existing technical solutions, there are three solutions: 1) only using the signals of the car itself to perform recursion, such as fitting the front wheel steering angle relationship using the steering wheel to give the heading of the car body, and then using the wheel speed to obtain the position of the car, so as to obtain the attitude of the car. However, in this solution, the accuracy of the steering wheel angle and the wheel speed signal is not high enough, and the fitting accuracy of the steering wheel to the front wheel steering angle also needs to be evaluated. Therefore, the overall accuracy is not high; 2) using IMU+wheel speed signals to perform fusion through Kalman filtering. The heading of the car is mainly given by the gyroscope of the IMU, and the wheel speed is used to correct the position, so as to give the attitude of the car. However, in this solution, the wheel speed signal transmitted by the car chassis has errors, which will affect the position and posture of the car. In addition, the car drives on different road conditions, and the load of the car is different. These factors will have a considerable influence on the accuracy of the odometer; 3) on the basis of solution 2, the wheel speed is corrected online through the GNSS signal. Therefore, the accuracy of the odometer will be improved accordingly. However, when the GNSS signal is weak or completely unable to be received, the correction of the wheel speed will not work.

[0004] Therefore, in the existing technical solutions, the wheel speed output is directly used to participate in the calculation of the odometer, without considering the influence of some external factors on the accuracy of the wheel speed, and it is necessary to obtain high wheel speed accuracy under the condition of no GNSS signal. SUMMARY

[0005] To address the aforementioned problems, the main objective of this invention is to design a wheel speed correction method for intelligent driving, solving the problem of wheel speed correction under different road conditions and variable loads, thereby improving the accuracy of local odometers.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A wheel speed correction method for intelligent driving, wherein the autonomous vehicle is equipped with an inertial navigation system, the method includes:

[0008] Step 1: Load the extrinsic parameters between the calibrated inertial devices and the vehicle center;

[0009] Step 2: Obtain the relevant signals of the inertial devices and wheel speed;

[0010] Step 3: Preprocess the inertial devices and wheel speed signals;

[0011] Step 4: Determine if the processed data is abnormal. If abnormal, return to step 2.

[0012] Step 5: Based on the wheel speed, classify and categorize the data without anomalies into speed ranges;

[0013] Step 6: Select the scene using the position information from the inertial devices and load the corresponding preset values;

[0014] Step 7: Select the load condition preset value through the vehicle load signal;

[0015] Step 8: Solve for the wheel speed correction model parameters;

[0016] Step 9: Generate the corresponding velocity scale coefficient lookup table.

[0017] As a further description of the present invention, the modules involved in the method include a parameter loading module, a data acquisition module, a data preprocessing module, a model fitting module, and a result generation module.

[0018] As a further description of the present invention, the specific steps of the method include:

[0019] Step 1: Load the extrinsic parameters between the calibrated inertial device and the vehicle center using the parameter loading module;

[0020] Step 2: Real-time acquisition of inertial device and vehicle wheel speed signals via data acquisition module;

[0021] Step 3: The data collected in Step 2 is first parsed using the data preprocessing module, and then synchronized and unified in terms of time and space.

[0022] Step 4: Evaluate the wheel speed and inertial device speed data processed in Step 3:

[0023] If the data is normal, proceed to the following steps; if the data is abnormal, return to step 2.

[0024] Step 5: Based on the wheel speed, classify and categorize the wheel speed and inertial device data that are normal after synchronization into speed ranges;

[0025] Step 6: Determine the scene using the position information from the inertial device and load the corresponding road condition factor K. r The expected value;

[0026] Step 7: Apply the load condition factor K using the vehicle weight signal. l The expected value;

[0027] Step 8: Solve for the wheel speed correction model parameters using the model fitting module;

[0028] Step 9: Finally, the corresponding velocity scale coefficient lookup table is generated through the result generation module.

[0029] As a further description of the present invention, in step 1, the loading parameters include: 1) the installation distance of the inertial device from the center of the vehicle; 2) the relative angle between the coordinate system of the inertial device and the coordinate system of the vehicle.

[0030] As a further description of the present invention, in step 2, the vehicle wheel speed, the inertial device speed, and the acceleration and angular velocity of the IMU in the inertial device are collected.

[0031] As a further description of the present invention, in step 3, the velocity of the inertial device is linearly interpolated by the timestamp of the inertial device to obtain the velocity of the inertial device at the corresponding wheel speed.

[0032] The specific interpolation method is as follows:

[0033]

[0034] Among them, v i，k Let v be the velocity of the inertial device at time k. i，k+1 Let t be the velocity of the inertial device at time k+1. i，k Let t be the timestamp of the inertial device at time k. i，k+1 Let t be the timestamp of the inertial device at time k+1. w v′ is the timestamp for wheel speed. i For the interpolated inertial device at t w The velocity at time t; satisfying: t i，k ≤t w <t i，k+1 ;

[0035] After time synchronization, spatial alignment is required using extrinsic parameters between the inertial devices and the vehicle center.

[0036]

[0037] in, This is the cumulative attitude change matrix of the IMU in the inertial device; This refers to the installation distance from the inertial device to the center of the vehicle. ω represents the angular velocity of the IMU at the corresponding moment; This refers to the velocity of the inertial device after spatial synchronization.

[0038] As a further description of the present invention, in step 4, the criteria for judging the wheel speed and the speed data of the inertial device are as follows:

[0039] The ratio r between the wheel speed and the speed of the inertial device is:

[0040]

[0041] Among them, v w v is the wheel speed. i Inertial device speed;

[0042] The data is normal if the ratio r ∈ [0.9, 1.1].

[0043] Data anomalies are: ratios r∈[0, 0.9) or r∈(1.1, +∞).

[0044] As a further description of the present invention, in step 8, a wheel speed correction model is established: Where v noise For velocity noise, K is solved using the least squares method. l and K r .

[0045] As a further description of the present invention, in step 8, the data pairs for each speed range are operated on as follows:

[0046] The matrix corresponding to wheel speed is:

[0047] V w =[v w,1 ... v w,N ] T

[0048] After time synchronization and spatial system integration, the matrix corresponding to the velocity of the inertial devices is:

[0049]

[0050] Error:

[0051]

[0052] Kl and K r This can be obtained by solving the least squares method:

[0053]

[0054] As a further description of the present invention, in step 9, after data processing for different speed ranges, the generated speed scale coefficient lookup table includes: speed and the corresponding road surface condition factor and load condition factor under different speed conditions.

[0055] Compared with the prior art, the technical advantages of the present invention are as follows:

[0056] This invention provides a wheel speed correction method for intelligent driving. By using a speed time coefficient calibration method under different road conditions and variable load conditions, it is applied to the IMU+wheel speed fusion odometer scheme to reduce wheel speed error and improve the accuracy of local odometer. In addition, by generating a speed scale coefficient lookup table, the computational requirements of the operating platform are lower, which lays a good foundation for the advancement of embedded schemes. Attached Figure Description

[0057] Fig. 1 This is a schematic diagram of the overall process of the wheel speed correction method of the present invention;

[0058] Fig. 2 This is a schematic diagram of the various modules involved in the wheel speed correction method of this invention. Detailed Implementation

[0059] The present invention will now be described in detail with reference to the accompanying drawings:

[0060] A wheel speed correction method for intelligent driving, reference Figs. 1-2 As shown, the system modules involved in this method include: parameter loading module, data acquisition module, data preprocessing module, model fitting module, and result generation module.

[0061] A detailed analysis of each module of the system involved in this method is conducted, and the analysis content is as follows:

[0062] I. Parameter Loading Module:

[0063] Establish a vehicle coordinate system and an inertial device coordinate system, with the positive directions being forward and left. That is, the forward direction of the vehicle is the positive x-axis, the left direction of the vehicle is the positive y-axis, and the upward direction of the vehicle is the positive z-axis. The center of the vehicle coordinate system is at the center of the rear axle of the vehicle, and the center of the inertial device coordinate system is at its geometric center.

[0064] Load the extrinsic parameters between the calibrated inertial device and the vehicle center, including the distance from the inertial device to the vehicle center and the installation angle.

[0065] II. Data Acquisition Module:

[0066] Speed ​​information containing inertial devices is collected in real time through inertial devices installed on the vehicle; wheel speed information of the vehicle is collected in real time through a CAN bus on the chassis.

[0067] III. Data Preprocessing Module:

[0068] The system analyzes real-time acquired inertial device information to obtain the required inertial device speeds; it also analyzes the vehicle chassis CAN bus to obtain the corresponding wheel speeds. The inertial device speeds are synchronized in time and spatially, and the system checks for anomalies and classifies speed ranges.

[0069] Data judgment, that is, data that is considered reasonable, such as: if the speed of the inertial device is 2km / h and the wheel speed is 0km / h, then it is rejected;

[0070] Speed ​​range classification: for example, if the resolution is 1 km / h, then data pairs of 0-1 km / h are grouped into one category, data pairs of 1-2 km / h are grouped into another category, and so on.

[0071] IV. Model Fitting Module:

[0072] The corresponding model is fitted using the least squares method.

[0073] V. Result Generation Module:

[0074] Generate a lookup table of velocity scale coefficients for different scenarios.

[0075] In this embodiment, based on the modules mentioned above, the specific steps of the wheel speed correction method for intelligent driving include:

[0076] Step 1: Load the extrinsic parameters between the calibrated inertial device and the vehicle center using the parameter loading module;

[0077] Step 2: Real-time acquisition of inertial device and vehicle wheel speed signals via data acquisition module;

[0078] Step 3: The data collected in Step 2 is first parsed using the data preprocessing module, and then synchronized and unified in terms of time and space.

[0079] Step 4: Evaluate the wheel speed and inertial device speed data processed in Step 3:

[0080] If the data is normal, proceed to the following steps; if the data is abnormal, return to step 2.

[0081] Step 5: Based on the wheel speed, classify and categorize the wheel speed and inertial device data that are normal after synchronization into speed ranges;

[0082] Step 6: Determine the scene using the position information from the inertial device and load the corresponding road condition factor K.r The expected value;

[0083] Step 7: Apply the load condition factor K using the vehicle weight signal. l The expected value;

[0084] Step 8: Solve for the wheel speed correction model parameters using the model fitting module;

[0085] Step 9: Finally, the corresponding velocity scale coefficient lookup table is generated through the result generation module.

[0086] Specifically, in this embodiment, the steps of the wheel speed correction method are analyzed in detail, and the analysis content is as follows:

[0087] Step 1: Load parameters, among which important parameters include: 1) Installation distance of the inertial device from the center of the vehicle. 2) Relative angle between the inertial device coordinate system and the vehicle coordinate system

[0088] In this embodiment, a rotation matrix is ​​used. This is used to represent the transformation between the inertial device coordinate system and the vehicle coordinate system. and The relationship between them is:

[0089]

[0090] Step 2: Collect vehicle wheel speed v wx Then, the wheel speed in the vehicle coordinate system is: Inertial device speed Acceleration of IMU in inertial devices and angular velocity

[0091] Step 3: Generally speaking, the frequency of the inertial device is higher than the frequency of the wheel speed. Therefore, in order to improve accuracy, the speed of the inertial device is linearly interpolated by the timestamp of the inertial device to obtain the speed of the inertial device at the corresponding wheel speed.

[0092] The specific interpolation method is as follows:

[0093]

[0094] Among them, v i,k Let v be the velocity of the inertial device at time k. i,k+1 Let t be the velocity of the inertial device at time k+1. i,k Let t be the timestamp of the inertial device at time k. i,k+1 Let t be the timestamp of the inertial device at time k+1. w v′ is the timestamp for wheel speed. i For the interpolated inertial device at tw The velocity at time t; satisfying: t i,k ≤t w <t i,k+1 ;

[0095] After time synchronization, spatial alignment is required using extrinsic parameters between the inertial devices and the vehicle center.

[0096]

[0097] in, This is the cumulative attitude change matrix of the IMU in the inertial device; This refers to the installation distance from the inertial device to the center of the vehicle. ω represents the angular velocity of the IMU at the corresponding moment; This refers to the velocity of the inertial device after spatial synchronization.

[0098] In step 4, the criteria for judging the wheel speed and inertial device speed data are: data that can be considered reasonable has a wheel speed to inertial device speed ratio within a certain range.

[0099] Among them, v w v is the wheel speed. i Inertial device speed;

[0100] The data is normal if the ratio r ∈ [0.9, 1.1].

[0101] Data anomalies are: ratios r∈[0, 0.9) or r∈(1.1, +∞).

[0102] Step 5: Classify the synchronized wheel speed and inertial device data based on the wheel speed magnitude;

[0103] For example, if the resolution is 1 km / h, then data pairs of 0-1 km / h are grouped into one category, data pairs of 1-2 km / h are grouped into another category, and so on.

[0104] Step 8: Establish the wheel speed correction model: Where v noise For velocity noise, K is solved using the least squares method. l and K r .

[0105] The solution to K is given below. l and K r Steps:

[0106] Perform the following operations on the data pairs for each speed range:

[0107] The matrix corresponding to wheel speed is:

[0108] Vw =[v w，1 ... v w，N ] T

[0109] After time synchronization and spatial system integration, the matrix corresponding to the velocity of the inertial devices is:

[0110]

[0111] Error:

[0112]

[0113] You can then obtain K l and K r Related least squares problems:

[0114]

[0115] Based on the solution steps of the least squares method, calculate the error term with respect to the state variable K. l and K r The derivative:

[0116]

[0117] This yields the Jacobian matrix:

[0118] The incremental equation for Gauss-Newton's method is: Where Δx=[ΔK l ΔK r ], for K during the iteration process l and K r The corresponding increment.

[0119] Specifically, the solution to this equation is as follows:

[0120] 1) Given K l and K r Initial value;

[0121] 2) For the k-th iteration, find the current Jacobian matrix J and error e;

[0122] 3) Solve the incremental equations of the Gauss-Newton method to obtain Δx;

[0123] 4) If Δx is sufficiently small, then stop, i.e., the required K is obtained. l and K r ;

[0124] Otherwise, let x k+1 =x k +Δx, return to step 2, and re-execute the operation steps of the wheel speed correction method.

[0125] Step 9: After processing the data for different speed ranges, the generated speed scale coefficient lookup table includes: speed and the corresponding road surface condition factor and load condition factor under different speed conditions.

[0126] The velocity scale coefficient lookup table generated above has the following format:

[0127] speed (km / h) Road condition factor K r ]]> Load condition factor K l ]]> 0～0.5 0.02123 0.0254 0.5～1.0 0.01892 0.02224 1.0～1.5 0.02003 0.02784 ... ... ...

[0128] It should be noted that the road surface condition factor and load condition factor values ​​corresponding to the speed ranges disclosed above are for reference only. In the specific implementation process, for different speed ranges, it is necessary to perform reasonable calculations based on the above overall steps to obtain the corresponding road surface condition factor values ​​and load condition factor values.

[0129] Based on this embodiment, a method for calibrating the speed scale coefficient under different road surface and variable load conditions is proposed for an odometer scheme that integrates IMU and wheel speed.

[0130] The technical solution of this embodiment reduces wheel speed error and improves the accuracy of local odometer. In addition, the final result of this embodiment is to generate a speed scale coefficient lookup table. Through calibration, the GNSS signal is fixed to the lookup table, which can meet the needs in the absence of GNSS. Since there is no online calibration calculation, computing power is saved, the computing power requirements of the running platform are relatively low, and it is more friendly to embedded systems. It is also applicable to the calibration of wheel speed accuracy in multiple scenarios and with variable loads, and the accuracy is greatly improved.

[0131] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Any other modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention, as long as they do not depart from the spirit and scope of the technical solutions of the present invention, should be covered within the scope of the claims of the present invention.

Claims

1. A wheel speed correction method for intelligent driving, used in autonomous vehicles, wherein the autonomous vehicle is equipped with an inertial navigation system, characterized in that: The method includes: Step 1: Load the extrinsic parameters between the calibrated inertial devices and the vehicle center; Step 2: Obtain the relevant signals of the inertial devices and wheel speed; Step 3: Preprocess the inertial devices and wheel speed signals; The velocity of the inertial device is obtained by linearly interpolating the inertial device's velocity using its timestamp; the specific interpolation method is as follows: ； in, Let k be the velocity of the inertial device at time k. Let k be the velocity of the inertial device at time k+1. Let k be the timestamp of the inertial device. The timestamp of the inertial device at time k+1. The timestamp for wheel speed. For interpolated inertial devices in The speed of time; satisfying: ; After time synchronization, spatial alignment is achieved using extrinsic parameters between the inertial devices and the vehicle center. ； in, This is the cumulative attitude change matrix of the IMU in the inertial device; This refers to the installation distance from the inertial device to the center of the vehicle. ; This represents the IMU angular velocity at the corresponding moment. This refers to the velocity of the inertial devices after spatial synchronization; Step 4: Determine if the processed data is abnormal. If abnormal, return to step 2. Step 5: Based on the wheel speed, classify and categorize the data without anomalies into speed ranges; Step 6: Select the scene using the position information from the inertial devices and load the corresponding preset values; Step 7: Select the load condition preset value through the vehicle load signal; Step 8: Solve for the wheel speed correction model parameters; Step 9: Generate the corresponding velocity scale coefficient lookup table.

2. The wheel speed correction method for intelligent driving according to claim 1, characterized in that: The modules involved in this method include a parameter loading module, a data acquisition module, a data preprocessing module, a model fitting module, and a result generation module.

3. The wheel speed correction method for intelligent driving according to claim 2, characterized in that: The specific steps of this method include: Step 1: Load the extrinsic parameters between the calibrated inertial device and the vehicle center using the parameter loading module; Step 2: Real-time acquisition of inertial device and vehicle wheel speed signals via data acquisition module; Step 3: The data collected in Step 2 is first parsed using the data preprocessing module, and then synchronized and unified in terms of time and space. Step 4: Evaluate the wheel speed and inertial device speed data processed in Step 3: If the data is normal, proceed to the following steps; if the data is abnormal, return to step 2. Step 5: Based on the wheel speed, classify and categorize the wheel speed and inertial device data that are normal after synchronization into speed ranges; Step 6: Determine the scene using the position information from the inertial devices and load the corresponding road condition factors. The expected value; Step 7: Apply load condition factor using vehicle weight signal. The expected value; Step 8: Solve for the wheel speed correction model parameters using the model fitting module; Step 9: Finally, the corresponding velocity scale coefficient lookup table is generated through the result generation module.

4. The wheel speed correction method for intelligent driving according to claim 3, characterized in that: In step 1, the loading parameters include: 1) the installation distance of the inertial device from the center of the vehicle; 2) the relative angle between the coordinate system of the inertial device and the coordinate system of the vehicle.

5. The wheel speed correction method for intelligent driving according to claim 3, characterized in that: In step 2, the vehicle wheel speed, inertial device speed, acceleration and angular velocity of the IMU in the inertial device are collected.

6. The wheel speed correction method for intelligent driving according to claim 3, characterized in that: In step 4, the criteria for judging wheel speed and inertial device speed data are as follows: The ratio of its wheel speed to the speed of the inertial device for: ； in, For wheel speed, Inertial device speed; The data shows no anomalies in the following cases: ratio ; The data anomaly is: ratio or .

7. The wheel speed correction method for intelligent driving according to claim 3, characterized in that: In step 8, establish wheel speed. Corrected model: ,in For velocity noise, the load condition factor is solved using the least squares method. and pavement condition factors .

8. The wheel speed correction method for intelligent driving according to claim 7, characterized in that: In step 8, the following operations are performed on the data pairs for each speed range: Wheel speed The corresponding matrix for: ； After time synchronization and spatial system integration, the matrix corresponding to the velocity of the inertial devices is: ； Error: ； and This can be obtained by solving the least squares method: 。 9. The wheel speed correction method for intelligent driving according to claim 3, characterized in that: In step 9, after processing the data for different speed ranges, the generated speed scale coefficient lookup table includes: speed and the corresponding road surface condition factor and load condition factor under different speed conditions.

Images