Research On Driving Risk Perception And Warning Method Based On IoV Basic Safety Message

The development of internet of vehicle(Io V)technology provides a new idea for the research of road traffic safety management application.In the lasted released Io V communication protocol standards SAE J2735-2016 and T/CSAE 53-2017,Basic Safety Message(BSM)related with driving safety can be collected in the Io V environment.In this thesis,BSM data is collected by real vehicles,and we use it to comprehensively perceive and predict the motion state of vehicles and dangerous driving behavior,and finally the collision prevention and warning research of vehicles in the Io V environment is realized.Firstly,we take the Nanchang-Jiujiang Intelligent Expressway test section of Jiangxi province as an example,and introduce the basic architecture and the key technologies of Io V technology and the classification standards of intelligent road.The driving safety data protocols are also sorted out.On this basis,combined with the Nanchang-Jiujiang Intelligent Expressway,a real vehicle test platform based on the Io V is built,and the real vehicle test is carried out to collect the BSM data.Secondly,in order to effectively improve the accuracy of BSM motion information,a BSM motion information collection method based on cooperative vehicle-road is established.Through the use of combination model of interactive multi-model filter and improved vehicle kinematics model,we combine the BSM motion status of self-vehicle collected by on-board unit and roadside unit,so the optimization of vehicle motion status estimation is realized when vehicle moves with large maneuvers.Compared with other method,and the result shows that the proposed model achieves less overall error and better tracking effect.What is more,in order to effectively perceive the driving status existing potential risks in the process of driving under the Io V environment,a method of distinguishing dangerous driving status based on rough set and improved support vector machine(SVM)is proposed.The rough set is used to select 8 factors from BSM as input of the model,and the combination model of genetic algorithm and SVM is used to optimize and calibrate the model parameters quickly.Compared with other methods,the result shows that this method is more accurately and can identify 94.44% of the potential risk driving status.Finally,to deal with the problem of uncertain driving behavior in complex road traffic scenarios,the deep learning theory is introduced into the research of situation awareness of vehicle forward collision risk.We fully consider driving scenarios like car following,lane change,free driving,and take continuous BSM data set as a sample,and build the multi-dimensional driving safety state sequence.We input the sequence into the LSTM neural network to predict the vehicle acceleration in a short time(0.5s).Then we use ROC curve to evaluate the predicted result,and use Youden index to select the optimal threshold value of potential collision risk based on the predicted result.Compared with other method,the result shows that the method could better detect collision risk and predict 87.838% potential forward collision risk events in 0.5s advance.In this thesis,we have carried out research of traffic risk perception and pre-warning method from the perspective of Io V.The BSM information obtained from the Io V is selected and effectively integrated from multiple perspectives,and we have achieved some research results.The research results have certain significance for improving the level of road traffic safety,and accelerating the development and application of Io V in the field of traffic safety.