Abstract
In the context of conditionally automated driving, drivers can engage in non-driving related tasks (NDRTs). However, it is imperative for them to resume control of the vehicle upon request, commonly referred to as takeovers. Despite the predetermined scheduling of takeovers, drivers may exhibit inadequacies in disengaging from ongoing NDRTs, thereby introducing safety risks. This thesis is dedicated to comprehensively understanding the dynamics of drivers’ disengagement from NDRTs during scheduled takeovers.This thesis started with a review of established definitions and theories relevant to the topic (Chapter 2) and a discussion on associated research methodologies (Chapter 3). Subsequently, Chapter 4 presents an empirical study that systematically observed participants’ disengagement from NDRTs, considering visual, cognitive, and physical dimensions. Furthermore, the study evaluates the impacts of varying NDRT disengagement timings on takeover performance and derives a pivotal concern that becomes the focus in subsequent studies—namely, incomplete cognitive disengagement from NDRTs (ICDN).
Then the thesis delves into the exploration of factors influencing ICDN through in depth interviews with drivers who underwent scheduled takeover experiments (Chapter 5). Following this, Chapter 6 presents the validation of the influencing factors, distinguishing between intentional and unintentional ICDN. Moving forward, Chapter 7 offers an analysis of objective measures of ICDN categories and details attempts to construct automatic classifiers using machine learning.
This thesis identified ICDN as a significant issue during scheduled takeovers, and provided insights for its prediction, intervention, and detection. These contributions advance the development of driver assistance systems in autonomous vehicles.
| Date of Award | Oct 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Qingfeng Wang (Supervisor), Xu Sun (Supervisor) & Jiang Wu (Supervisor) |