Research On Safety Test Method Of Take-Over Signal For Autonomous Vehicle With Comfort

As an important technical problem for the commercial use of autonomous vehicles,human-machine takeover has become a worldwide concern.How to ensure the safety and take into account the comfort of the driver,so as to avoid accidents and take into account the usability of autonomous vehicles has become an important problem to be solved urgently.In this paper,the take-over warning signals of autonomous vehicles are studied from the aspects of safety,comfort,boundary acquisition of test scenes and simulative degree of test tools,etc.A set of extensible human-machine take-over safety test method with comfort in typical scenes is formed.The main research results are as follows:The quantitative characterization of driver comfort by physiological indexes is studied.The typical boundary scenarios of autonomous driving were simulated by autonomous driving robots in real vehicles,and a closed-field experiment was conducted with 9 tested drivers using a two-factor experiment.Heart rate,heart rate variability time domain indices SDNN(Standard Diviation of NN Intervals)and RMSSD(Root Mean Square of the Successive Differences)and Frequency domain indicators LF(Low Frequency)and HF(High Frequency)are extracted and calculated.The effects of different motion states on the above parameters were analyzed by using repeated measurement general linear model,and driver comfort was analyzed from both objective and subjective aspects.It was confirmed that the discomfort caused by braking would lead to a slight increase in heart rate,a significant decrease in SDNN and RMSSD,and different horizontal accelerations had a significant impact on the change degree of SDNN.It is confirmed that there is a significant correlation between physiological indicators and driver comfort and automatic driving scenes,and the subjective evaluation of comfort is highly consistent with the representation of physiological indicators data.The research results prove that it is feasible to objectively represent the driver comfort of autonomous vehicles in specific scenes based on physiological indicators.The fidelity and applicability of driving simulator are studied.In the dynamic following scene,the driving performance and subjective evaluation of drivers in driving simulator were compared with the objective index of following distance and the objective driving performance and subjective evaluation of actual vehicles,and the fidelity of driver operation in driving simulator was studied.A closed field experiment was conducted on 10 drivers using two two-factors experiment.The influences of 2D and S3 D viewing and whether the motion platform function is enabled on drivers’ following performance in driving simulator are studied.Meanwhile,the subjective feelings are interviewed and the applicability of driving simulator is investigated.The study confirmed that S3 D can provide more depth information,provide better immersion experience for the subjects,and create more realistic environment perception and distance judgment,thus having an accuracy advantage of 2.5m(95%CI: 1.11-3.85m)compared with 2D mode.It is clear that under the general driving dynamic task,there is no statistical difference between the driving performance of the driving simulator and the real driving situation,and it meets the requirement of realism for man-machine takeover test.The method for determining test boundary conditions in typical human-machine takeover scenarios is studied.Based on the theoretical research of the scene,the definition of the scene was clarified,the real driving data acquisition platform was built,and the natural driving data of 1000 km was collected.Four typical driving scenarios were extracted from random natural driving behaviors by setting different extraction rules for the scenes.Based on the analysis of characteristic indexes such as headway,headway and speed and prior knowledge judgment,the cutting scene of the vehicle in front was selected as the limit scene,and the boundary test conditions based on the driving simulator were studied based on 12 tested persons.Adjust TTC(Time to Collision),vehicle speed,relative speed and other traffic participant parameters to approach the safety critical value,and finally determine the ultimate safety critical condition TTC=1.5s in the cut-in scenario.A set of safety testing methods for alarm signals of take-over considering comfort in typical scene limit boundary conditions is proposed.The test method was based on the confirmation of limit boundary conditions of typical scenes,take-over test of driving simulator,physiological index collection and subjective and objective analysis,and finally completed the safety test and comfort evaluation analysis of audible and visual take-over warning signals under the limit boundary conditions of typical scenes.This set of methods has theoretical basis and expansibility.It can extend the safety test considering comfort at more physiological signals,multiple typical scenarios and multiple take-over prompts,thus providing strong support for product design and development and product access.