Lane-Changing Behavior Based On Eye Movement Characteristics In Rainy Environment

Bad weather is one of the causes of traffic accidents.In order to study the relationship between bad weather and vehicle lane change,improve vehicle driving safety factor and improve active safety prevention and control measures,this paper carries out real vehicle test and simulated driving test with the help of Dikablis Glass 3eye tracker,D-Lab software and simulated driving test room.Collect and analyze the driver’s driving eye movement data and vehicle operation parameters in the rainy environment,explore the correlation between the driver’s eye movement parameters and vehicle parameters and lane changing behavior,establish the vehicle lane changing model in rainy days,and put forward the method of lane changing risk early warning.The specific work is as follows:(1)According to the existing research results at home and abroad and actual needs,determine the type of lane change,lane change stage,and starting point of lane change intention studied in this paper,and design the actual vehicle test plan and simulated driving test plan.Four screening conditions were proposed and 144 effective lane changing processes were screened.After eliminating the abnormal data in the lane change data,K-means algorithm is used for fixation clustering,which lays the foundation for the follow-up eye movement research.At the same time,the vehicle operation parameters in rainy environment are compared and analyzed.The parameters mainly include driving speed,approximate entropy of steering wheel angle and driver’s use rate of turn signal.(2)Study the eye movements of drivers in the process of changing lanes on rainy days.According to the actual situation of this article,the eye movement data of drivers is compared from four aspects of different weather and different rainfall,6 areas of interest,3 lane changing phases and different proficiency of drivers.Combining the Markov state transition theory and software analysis,the gaze probability and transition probability of the driver in rainy days under different rainfall are obtained.At the same time,according to Markov stable distribution characteristics,under different rainfall environments,when the driver’s gaze point reaches a stable state after multiple path shifts,the gaze probability of the 6 regions of interest is studied.(3)Establish a rainy lane change model and conduct a rainy lane change risk analysis.Quantify the data of two aspects: the law of driver’s eye movement parameters and the law of vehicle operating parameters.Including rainfall,vehicle lateral acceleration,approximate entropy of steering wheel angle,number of glances,stable gaze probability of target lane,current lane-left area gaze transition probability,current lane-right area gaze transition probability,and current lane-target lane gaze transition probability.Logistic regression analysis and mixed Gaussian Hidden Markov theory are used to establish Logistic regression model and GMM-HMM model respectively.After verification,the accuracy of Logistic regression model is only 62.2%,and the total recognition accuracy of GMM-HMM model can reach 90.46%.This proves that the GMM-HMM model can accurately identify lane changing behavior.At the same time,improve the minimum distance model of vehicles and establish the conflict probability model,formulate the risk identification and early warning mechanism of lane change,conduct risk analysis of lane change in rainy day,and improve the active safety prevention and control measures in rainy day.