A Study Of The Effectiveness Of Automatic Driving Takeover Task Based On Driving Simulation Test

In level 3 conditional automatic driving,there are still many operations that the automatic driving system cannot complete,so it has to hand over the control of the vehicle to the human driver to reduce the driving risk,and ensure driving safety.However,the safety risk in the process of driver takeover cannot be ignored,and the effectiveness of the driver in performing the automatic driving takeover task needs to be further investigated.The main content of this paper is to study the process and results of drivers performing takeover tasks in different takeover scenarios under autonomous driving mode,explore the behavioral characteristics of drivers performing takeover tasks in different dangerous takeover scenarios,and construct a driver’s takeover effectiveness discrimination model,After obtaining control of the vehicle,the driver should make a judgment on the effectiveness or invalidity of the takeover task as soon as possible.The specific research content is as follows:(1)Define the basic concepts related to the effectiveness of driver takeover,such as takeover tasks.Firstly,based on the switching of vehicle control rights,the takeover task was defined,the start and end times of the takeover task were clarified,and the results of the takeover task were accurately defined as an effective and ineffective takeover.Then,the factors influencing the effectiveness of takeover were analyzed from both deterministic and non-deterministic perspectives,and a set of indicators for determining the effectiveness of driver takeover was constructed,providing a theoretical basis for the subsequent establishment of a model for determining the effectiveness of driver takeover.(2)Analyze the relationship between takeover behavior and the outcome of takeover tasks.Firstly,by designing and conducting autonomous driving takeover simulation experiments,all driving behavior data of drivers during the takeover task were obtained.Based on different behavior data,corresponding indicators for determining the effectiveness of driver takeover were extracted.Statistics were conducted on the driver’s takeover methods,the proportion of ineffective takeover,and the distribution of takeover task duration.By comparing the differences in indicators related to driver takeover response between different takeover scenarios and different takeover results within the same takeover scenario,it can be understood whether each indicator is beneficial for improving the accuracy,accuracy,and recall of the takeover effectiveness discrimination model.(3)Construction of a discriminant model for the effectiveness of driver takeover.This article does not distinguish between the degree of danger in takeover scenarios and the degree of danger in takeover scenarios.Different takeover effectiveness discrimination models are constructed by combining the discrimination indicators obtained by the driver at different times during the takeover task with multiple classifiers.Finally,the optimal takeover effectiveness discrimination model was determined by comparing different discrimination models’ accuracy,precision,and recall.This model realizes the discrimination of takeover task results based on the early stage of the driver executing the takeover task.