Event

Alberto Morando will defend his thesis "Drivers' response to attentional demand in automated driving" on March 21, 2019. Alberto is a PhD student at Chalmers, Mechanics and Maritime Sciences, Vehicle Safety. The opponent is Prof. John Lee from University of Wisconsin. A light lunch will be served after the defense. If you want to join the lunch, please let Sonja (sonja.laakso@chalmers.se) know by March 14^th.

Welcome!

Abstract

Introduction
Vehicle automation can make driving safer, by compensating for human impairments that are frequently recognized as the leading cause to crashes. Vehicle automation has become a central issue in transportation research, and there has been extensive research on the human factors of automated driving. There are open challenges to be addressed to understand how to guide attention for safe use of automation and how to improve the design of automated systems to accounts for human abilities and limitations. For example, what are the effects of automation on driver’s attention to the driving task? What are the safety implications of those changes? Are drivers prompt to respond to critical situations? How can attention be elicited in situations that require an intervention by the driver? Additionally, as the existing body of research is often based on simulator studies, the real-world effects of automation remain unclear.

Objectives
The thesis investigated how driver’s attention changes with automation and the driving situation—primarily in real-world driving. The objective was to inform the design of real-time systems and develop design knowledge to support safe driving. To pursue this objective, the thesis provided new evidence and proposed new methods for interpreting driver behavior.

Methods
The thesis leveraged on naturalistic driving data (to study the effect of assistive automation on driving behavior) and on data collected with a driving simulator experiment (to study the effect of highly automated driving). Driver behavior was examined with measures of visual and motor response (at aggregate level or across time) together with contextual information on the driving situation (vehicle and on the surrounding environment).

Results
The findings show that assistive automation (adaptive cruise control and lane keeping aid) affects driver’s attention in real-world driving. In general, drivers devoted less attention at the forward path when using low-automation compared to manual driving. However, driver’s visual response was sensitive to changes in the surrounding context (presence of other traffic and illumination) and it was elicited by perceptual cues (visual, audio, and vestibular-kinesthetic-somatosensory) that alerted of an impending conflict. The driver’s response chain to a critical situation with high automation had a reflexive (glance on-path, hands on wheel, and feet on pedals) and a voluntary component (execution of evasive maneuver). Moreover, the results indicate that expectation, which changes over time depending on experience, affected driver’s response substantially.

Conclusions
First, regarding the understanding of driver behavior and its assessment, the thesis shows that the influence of driving contextual factors is essential to understand behavioral changes and safety consequences in automation. Driving context helps to distinguish if attention is away from the forward path or from a potential threat (e.g., a lead-vehicle). Driving context depends itself on automation, which has the added benefit in reducing the exposure to critical situations compared to manual driving (e.g., by maintaining a safe headway to other traffic). Driving context is dynamic, therefore by investigating the time-course of attention with respect to critical events, one can understand if drivers are receptive, and prompt to respond, to cues (e.g., visual and vestibular-kinesthetic-somatosensory) that signal a threat. Second, the results from the empirical research improve on those of earlier studies (both experimental and on-road) by providing a comprehensive assessment of driver attentional response in routine driving and critical situations. The results are provided in a format to be used as a reference to compare with manual driving within driver modelling and assistance/automated system development. Further, they can be used for evidence-based recommendations (e.g., inattention guidelines). In fact, the thesis pursued alternative methods for data analysis and statistical modelling (e.g., Bayesian methods), which offers a richer interpretation of the data than traditional methods, encourages replication, and eases comparison across studies.