Research On Take-Over Performance Of Autonomous Vehicles Under Typical Urban Road Driving Scenarios

The automatic driving technology is developing rapidly with fierce momentum,but the current automatic driving technology has not been able to realize the full automatic driving of vehicles.The human-machine cooperative driving mode between manual driving and full automatic driving will still exist for a long time.In the environment of human-machine cooperative driving,the safety problem in the take-over process has become the focus of domestic and foreign scholars.Many scholars study the take-over performance of automatic driving from two aspects:take over time and take over quality,and explore the influence of various factors on the take over performance,which is of great significance to promote the application of automatic driving technology.Based on the autopilot simulation platform,this study carries out the autopilot takeover experiment,and makes qualitative and quantitative analysis on the takeover performance of autopilot vehicles.Firstly,based on the review of the domestic and foreign research on the performance of automatic driving take over,this paper summarizes the shortcomings of the existing research,extends the index analysis of the performance of automatic driving take over to the evaluation level,and then gives the research content and technical route of this paper.Secondly,based on the classification standard of automatic driving,this paper introduces the concept of human-machine cooperative driving and the classification of control right switching,summarizes the process of automatic driving taking over under human-machine environment,and classifies the influencing factors of taking over performance from two aspects of driving process risk and original risk,so as to provide a theoretical basis for subsequent analysis.Then,the performance evaluation model of autopilot is constructed.This paper defines the concept of take over performance of automatic driving vehicles,selects the total take over time,longitudinal acceleration,lateral acceleration and the reciprocal of collision time as the evaluation indexes from two aspects of take over time and take over quality,establishes the take over quality evaluation system,and uses AHP and improved entropy method to evaluate the evaluation indexes objectively,subjectively and combined weighting,fuzzy comprehensive evaluation method is selected to comprehensively evaluate the performance of autonomous vehicle takeover.Then,the autopilot simulation platform is selected and the experimental scheme is designed.Three typical urban road driving scenarios(such as intersection straight,section straight and intersection turning)and three weather conditions(such as sunny day,rainy day and snowy day)are selected to carry out the autopilot takeover experiment under different weather conditions in typical road driving scenarios,and the experimental data are collected,As the data source of subsequent automatic driving vehicle take over performance index extraction and analysis and evaluation.Finally,based on the index weight vectors obtained by different index weighting methods,the results of performance index analysis and combination evaluation under each experimental scenario are obtained respectively.The results of performance index analysis of automatic driving take over show that the take over time of the section straight scene is about 0.2S longer than that of the intersection scene,and the longitudinal speed control is more stable than that of the intersection scene,but the transverse control of the intersection turning is better than that of the intersection straight scene and the section straight scene;The influence of weather factors on the maximum longitudinal acceleration and the maximum lateral acceleration is the most significant,but the influence on the total take over time and the minimum TTCI is not significant.The longitudinal acceleration in rainy and snowy days is 1～2m/s~2smaller than that in sunny days,and the lateral acceleration in the intersection straight is about0.2m/s~2smaller than that in sunny days,the lateral acceleration in the section straight scene is about 0.25m/s~2smaller than that in the sunny day,and there is no significant difference in the lateral acceleration in the turning scene.The combination evaluation results show that the performance evaluation values based on different index weight vectors are different,but on the whole,the performance of intersection turning scene is the best and intersection straight scene is the worst;Compared with the weather conditions,the takeover performance is the best in sunny days and the worst in snowy days.