Research Based On Datadriven Methods For Carcollisionavoidance

In recent years,with the rapid development of science and technology,automatic driving technology is also developing rapidly.As one of the core techniques in automatic driving system,vehicle collision avoidance strategy has attracted more and more attention and become a research hotspot.On the other hand,with the situation of vehicle driving safety becoming more and more serious,it is very important to study vehicle collision avoidance system to improve driving safety.The traditional vehicle collision avoidance technology mainly includes two kinds: the first one to provide is accident warning by measuring the distance between vehicles and calculating travel time interval.However,this method can not meet the real time.The second one is to establish the mathematical model of vehicle moving system,and then calculate the braking distance and advised acceleration.But because of complex characteristics of the vehicle,the built model cannot accurately reflect the motion characteristics of the vehicle.There is the problem of poor accuracy.In order to solve above problems and achieve the purpose of vehicle collision avoidance,it has become a new solution to study the new effective collision avoidance technology.In recent years,big data technology develops rapidly.And the massive data generated by human-computer interaction,often contains a lot of information that can reflect the system running state and even the key information of the system operation mechanism.At the same time,with the advent of the Industrial 4.0 era,the Cyber Physical System(CPS)has become a hot topic.It can combine material resources and computing resources,which is beneficial to achieve intelligent control and adaptive control.Considering these advantages,this paper combines the data-driven technology and CPS technology.Its aim is to provide warning information and advice acceleration to achieve the purpose of vehicle collision avoidance.This paper mainly contains three research contents:the modeling of vehicle cyber physical system(VCPS),the research of data-driven prediction algorithm,the simulation experiments.First of all,this paper studies the modeling of the VCPS,designs the structure and the mathematical modeling process.Secondly,this paper studies the theory and application of data driven algorithms,including multivariate statistical analysis,principal component analysis,partial least squares regression,kernel partial least squares regression and just in time learning.According to the different application characteristics,the paper focuses on the three data-driven methods,i.e.PLS,KPLS and JITL-PLS and proposes an improved JITL algorithm of JITL-PLS.Thirdly,the simulation experiments is operated in the built VCPS simulation platform.This experiment designed the test environment on the Car Sim and inspected the effectiveness of these data-driven methods.The experimental results show thatthe data driven algorithm based on JITL-PLS has higher prediction accuracy and shorter use time,which is suitable for the research purpose of this paper.Finally,this paper summarizes the content of this study and looks forward to the next step.