High Precision Localization Based Connected Vehicle Rear End Collision Avoidance Theory And Its Safety Evaluation

Vehicle technologies have achieved significant breakthroughs in the last few years,including humanless driving technology and intelligentiallization of vehicle.The safety issues of intelligent vehicle are becoming more and more important,one of the most momentous vehicle safety aspects is rear end collision avoidance.Current widely applied rear end collision avoidance systems are mainly based on vehicle onboard sensors to acquire information of surrounding vehicles,passengers and environments.Those vehicle sensors are easily interfered by rain,snow,fog and haze,and on the other hand,their applications are restricted by obstracts and limited detection ranges,consequently,vehicle sensor based collision avoidance systems cannot guarantee vehicle safety in all situations.With the rapid developments of high precision localization methods and communication techniques,high precision vehicle localization based connected vehicle active safety technologies have aroused more and more attentions all over the world,and this field is one of recently emergent industries that should be developed for our country.The present dissertation relys on National High-tech Research and Development Plan(863 Project)“Urban Vehicle Online Location Service Technology for Active Traffic Safety” and “Development and Demonstration Application of Decimeter-level Phase-enhanced Operation Service System for Central and Eastern China”,focuses on enhancing the rear end collision avoidance safety of intelligent vehicles,studies emergency braking and emergency steering based rear end collision avoidance algorithms,based on the former research,investigates stacked autoencoder neural network based rear end collision avoidance method,as well as risk evaluation approach for chain collision.To overcome the disadvantage that assisted GNSS based high precision localization method cannot be used in all environments owing to the inherit vulnerability of GNSS,this dissertation discusses 5G and Locata based vehicle localization methods,proposes SRCKF(Square Root Cubature Kalman Filter)based 5G and vehicle kinematics integrated high precision vehicle localization method,and apply this method into rear end collision avoidance algorithm research.The three primary research aspects of the present dissertation and corresponding innovations are summarized as follows:· Research on high precision localization based connected vehicle rear end collision avoidance theories,including emergency braking and emergency steering based rear end collision avoidance algorithms,as well as stacked autoencoder neural network based method.Investigate connected vehicle rear end collision avoidance algorithm when considering measurement uncertainties of vehicle localization,and the required safe distance under predefined safety probability is initially mathematically formulated as the rear end collision avoidance safety index,and safe distance margin and quantized influence two indexes are proposed to evaluate the influence of vehicle localization measurement uncertainties.These indexes can also allow for a better research and evaluation of rear end collision avoidance methods.Then,emergency braking and emergency steering two strategies based rear end collision avoidance methods are analyzed.Design stacked autoencoder neural network neural network based required safe distance prediction algorithm,which can replace the large amount of integration calculation when utilizing vehicle dynamics,and output real-time required safe distance prediction value under current vehicle and environment,and avoiding rear end collision.This is a new possible technology roadmap to apply Artificial Intelligence(AI)into rear end collision avoidance and vehicle safety applications.· Research chain collision risk evaluation method,evaluate the influence of device MPR(Market Penetration Rate)in reducing chain collision risk.Propose chain collision risk evaluation method,and considering that intelligent vehicles and traditional human driving vehicles will share roads in the future,design device MPR influence evaluation approach to chain collision risk.This proposed chain collision risk evaluation approach can be used to analyze the influence of road traffic flow,inter-vehicle communication delay,distribution of vehicle maximum decelerations and device MPR to chain collision risk,and construct a useful evaluation framework to be used in chain collision risk evaluation under mixed environment of intelligent vehicles and human driving vehicles.· Research 5G and Locata based vehicle localization methods and their application in rear end collision avoidance,evaluate corresponding performance.To overcome the restrictions on GNSS based vehicle localization because of GNSS vulnerability,investigate 5G and Locata based vehicle localization methods in order to replace GNSS.Utilizing SRCKF based 5G measurement and vehicle kinematics integrated high precision vehicle localization algorithm,by applying this high precision vehicle localization method into rear end collision avoidance algorithm,vehicle safety can be improved under all environments,including GNSS-denied situations.