A vehicle motion model optimization method based on deep learning

Abstract

The invention relates to a vehicle motion model optimization method based on deep learning, which fits a non-linear function between a steering wheel rotation angle alpha and a mass center deflectionangle beta in a deep learning manner, and finally obtains a process that the steering wheel rotation angle alpha is known and the mass center deflection angle beta is solved. A large amount of steer wheel information is collected through build a deep learning neural network training model, vehicle speed, centripetal acceleration and centroid deflection information in the vehicle motion model, thenonlinear correspondence between steering wheel angle and center-of-mass deflection angle is found, so as to optimize the steering wheel angle information in the vehicle motion model, to obtain more accurate information of motion direction, and to further improve the accuracy of the position estimation of the motion model.

Description

Technical field [0001] The invention relates to the technical field of intelligent driving, in particular to a method for optimizing a vehicle motion model based on deep learning. Background technique [0002] When using a simple vehicle motion model, the yaw angle θ cannot accurately describe the yaw angle of the true vehicle motion. Simple vehicle motion models such as figure 1 As shown, [0003] [0004] Among them, θ is the yaw angle, which is the counterclockwise angle relative to the X axis; [0005] v is the velocity in the θ direction; [0006] (x, y) are the position coordinates of the vehicle. [0007] The real vehicle kinematics model, such as figure 2 As shown, [0008] [0009] Among them, ψ is the yaw angle of the vehicle; β is the deflection angle of the center of mass of the vehicle. [0010] If the β centroid deflection angle can be obtained, the real vehicle kinematics model can be used to solve the accurate position estimation of the vehicle. The conventional method ...

AI Technical Summary

Problems solved by technology

The conventional method is to calibrate the steering wheel to detect the front tire deflection angle every 10°, and then use the calculation formula to solve the β centroid deflection angle. There are two problems caused by this method. The first is calibration when the vehicle is stationary, and the data acquisition is discrete. Yes, the intermediate difference needs to be obtained by integral, and the error is relatively large; the second is that the β centroid deflection angle is related to vehicle speed, centripetal acceleration, and tire cornering force. It is a continuous dynamic process, and it cannot be solved correctly from the static calibration method alone. Out of β centroid deflection angle

Images