An Examination of Drivers’ Responses to Take-over Requests with Different Warning Systems During Conditional Automated Driving

Abstract

Today, the autonomous vehicle industry is growing at a fast pace towards Level-5 autonomous cars, based on the Society of Automotive Engineers (SAE) definition, for customers. It is expected that there will soon be SAELevel-3 automated cars in the market–which corresponds to a plethora of research works in this sector and one of them is the study of the design of takeover request warning system because failure to respond a takeover request warning may lead to fatal accidents. The objective of this study is to examine the effects of different warning types on drivers’ takeover responses while they are engaging in different non-driving tasks during conditional automated driving. This study is a simulator-based with a mixed-subjects design while participants interacting with a simulated Level-3 automation system under different conditions. A total of 24 participants were recruited and participated in the study. Each participant experienced two types of takeover request (TOR )warning systems (Auditory TOR and Multimodal TOR) under four types of non-driving task conditions with two levels of non-driving task duration. One baseline drive without any secondary task was also designed for comparison with those conditions with non-driving tasks. Three research questions are addressed in this thesis: •Will a Multimodal TOR lead to better driver responses in reaction to takeover requests than Auditory TOR? •Will the different type of non-driving tasks lead to different cognitive engagement of drivers, therefore resulting in different reactions to takeover requests? Will different duration of engagement in non-driving tasks impact on responses of drivers’ re-engagement in driving tasks? In this study, data was collected for both objective driver measures through simulator run log files and subjective driver measures through questionnaires. For analysis purposes, a Mixed-Effects Model was conducted to test the response variables, followed by the Fisher LSD Pairwise Comparison test for significant factors with more than two levels and Two-Sample t-tests for subjective measures were used. Results showed that Multimodal TOR leads to shorter brake time and steer touching time comparatively and the difference of these dependent variables between the TORs is significant as p-value<0.05. The findings also suggest that the Multimodal TOR warning system leads to a better reaction of drivers. Moreover, it was also found that the type of non-driving tasks leads to different driver responses, more specifically, drivers have a significantly slower reaction towards the takeover request if they are engaging in visual-manual non-driving tasks when compared to if they are engaging in other types of non-driving tasks (e.g., cognitive or visual tasks). However, there are no significant gender-based effects observed for Brake Time and Steer Touch Time.