Method and system for previewing safe path based on CNN and LSTM

Abstract

The invention relates to a method and a system for previewing a safe path based on a CNN (Convolutional Neural Network) and an LSTM (Long Short Term Memory). The method at least comprises the following steps: predicting an ideal steering wheel rotation angle of an initial driver following model based on expected vehicle speed information and expected path information; optimizing the ideal steeringwheel rotation angle into an actual steering wheel rotation angle based on vehicle motion information fed back by a vehicle model; and optimizing vehicle attitude information based on the actual steering wheel rotation angle information and fitting to obtain expected path information by the vehicle model. According to the method and the system, through a preset road track and expected vehicle speed information, the steering wheel rotation angle applied to the vehicle by the driver is decided, so that the driving path of the driver is decided. and internal weight of a neural network is continuously adjusted through LSTM, so that deviation between an expected path and an actual path is reduced, and a good driver preview following model for preview path following is obtained.

Description

technical field

[0001] The present invention relates to the technical field of intelligent driving, in particular to a method and system for previewing safe paths based on CNN and LSTM. The invention also relates to the technical field of automobiles, and relates to an automobile that previews a safe path. Background technique

[0002] Path tracking is one of the cores of the underlying control of autonomous driving, and its performance determines the safety and stability of the vehicle. In the field of path tracking research, the more commonly used methods are: PID, driver preview model, model predictive control (Model Predictive Control, MPC) and so on. Among them, the PID algorithm has a simple structure, but it is difficult to adjust the parameters, and the fixed parameters cannot be used for complex and changeable driving conditions. Model predictive control can cope with different working conditions, but it has high requirements for the vehicle model and a large amoun...

Images