Study On Vehicle Collision Avoidance Strategy Based On Collision Warning And Deep Learning

Vehicle collision accidents bring great harm to passengers.Collision avoidance system which can effectively reduce the occurrence of collision accidents is also an important part of advanced driving assistance system(ADAS).However,the existing collision avoidance system usually only has the function of monitoring and early warning.It cannot predict the damage degree of vehicles and passengers in the future,and then make decisions.Aiming at the above problem,this paper develops a predictable intelligent car collision avoidance strategy based on real-time collision warning algorithm and deep learning.At the same time,it builds an intelligent car platform and uses binocular vision technology to verify the above collision avoidance strategy.The main contents of this paper are as follows:(1)The intelligent car collision avoidance strategy based on real-time collision avoidance warning algorithm and deep learning is studied.The Kalman filter algorithm is used to predict the motion state parameters of obstacles.Based on collision warning algorithm,the degree of collision damage under various schemes is calculated,and the optimal collision avoidance strategy is formulated.This intelligent car collision avoidance strategy is implemented by the algorithm.(2)The platform of intelligent car is designed and built.Select the modified platform,design the speed control part,the driving part,the sensor part and the control part and select the hardware device that meets the requirements of each part.Considering the power transmission and signal transmission of the car,design the hardware circuit is designed to control driving action such as driving,braking and steering According to the factors of stability and ride comfort,the structure of intelligent car is reasonably arranged,and the hardware construction of intelligent car is completed.(3)The principle of ranging,related algorithms,data acquisition and processing methods of binocular vision system are studied.Build outdoor experimental scenarios to verify the effectiveness of the collision avoidance strategy based on real-time collision warning algorithm and deep learning.Compared with the traditional intelligent car collision avoidance control strategy,our method can real-time predict the collision damage of intelligent car in a short time according to dozens of different driving strategies.And based on these predictions,the vehicle driving strategy corresponding to the least damaged is selected,and the driving behavior of the intelligent car is corrected in time to avoid the collision accident or reduce the collision damage.The development method of this paper is a new vehicle collision avoidance strategy,which is expected to be applied to real vehicles in the future,so it has important theoretical and application value.