Scenario Fidelity and Perceived Driver Mental Workload: Can Workload Assessment be Crowdsourced?
Abstract
Driver workload assessment is difficult. In spite of decades of effort, there is still no generally accepted method of assessing driver mental workload. Prominent methods include physiological, subjective, and dual-task assessments. While driving workload may also be influenced by the driving experience and mood, or other driver-specific factors, we were interested in the question of how much mental workload was likely to be induced in different types of driving situations. Since construction of driving situations in simulators is time consuming and may be limited by the technical features available, we have explored the use of videos of driving situations as stimuli for estimating driver mental workload. In the study reported here we asked two questions. 1) How well do people agree with each other when rating the mental workload that would be elicited by different driving situations presented as videos or storyboards? 2) How well do mental workload judgments based on storyboards of driving situations agree with corresponding judgments made after viewing videos of the same scenarios? We discuss the implications of this work for crowdsourcing driver workload assessment and building large data repositories suitable for training machine learning models. Our ultimate goal is to develop safe driver notification systems that schedule notifications when the driver is experiencing low (but not high) levels of mental workload.
Authors: Haoyan Jiang, Sachi Mizobuchi, Mark Chignell