Humans and Intelligent Vehicles: The Hope, the Help, and the Harm

Abstract

Abstract-Intelligent vehicles offer hope for a world in which crashes are rare, congestion is reduced, carbon emissions are decreased, and mobility is extended to a wider population. As long as humans are in the loop, over a half century of research in human factors suggests that this hope is unlikely to become a reality unless careful attention is paid to human behavior and, conversely, the potential for harm is real if little attention is given to said behavior. Different challenges lie with each of the two middle levels of automation which are the primary focus of this article. With Level 2 automation (National Highway Traffic Safety Administration; NHTSA), the driver is removed from always having to control the position and speed of the vehicle, but is still required to monitor both position and speed. Humans are notoriously bad at vigilance tasks, and can quickly lose situation awareness. Moreover, even if vigilant, the driver needs to interact with the vehicle. But voice-activated systems which let the driver continue to glance at the forward roadway are proving to be a potential source of cognitive distraction. With Level 3 automation (NHTSA), the driver is out of the loop most of the time, but will still need to interact with the vehicle. Critical skills can be lost over time. Unexpected transfers of control need to be considered. The surface transportation and aviation human factors communities have proposed ways to solve the problems that will inevitably arise, either through careful experimentation or extensions of existing research.

Deep CNN, Body Pose, and Body-Object Interaction Features for Drivers’ Activity Monitoring

“…We have submitted our results to the Kaggle and appeared on the top 17% in the leaderboard. The winner 1 used empirical tricks such as focusing head and driver's right hand 1 https://www.kaggle.com/c/state-farm-distracted-driver-detection/ discussion/22906 regions for fine-grained representations resulting in higher recognition accuracy. We would like to emphasise that our approach using the proposed high-level feature is simple yet effective and semantically rich.…”

“…It offers our desire for a better world in which injuries and fatalities from accidents are rare, congestion is lesser, and many societal and environmental benefits are far greater. It is suggested that this desire is unlikely to become a reality unless mindful attention is paid to human behaviour [1] since human error is overwhelmingly to blame for the vast majority of automobile accidents [2]. Today's automobile is nearly autonomous due to Advanced Driver Assistance System (ADAS) and is feasible due to ultrasonic sensors, cameras, radars and lidars.…”

Driver error or designer error: Using the Perceptual Cycle Model to explore the circumstances surrounding the fatal Tesla crash on 7th May 2016

“…The NTSB recognises that decades of research has shown that humans are notoriously inefficient at monitoring automated systems (e.g. Casner & Schooler, 2015;Fisher et al 2016;Molloy & Parasuraman, 1996) and provide a number of recommendations in an effort to improve system transparency and safety. However, the report still fails to explain why the driver felt able to relinquish control in the way that they did.…”

Latent Body-Pose guided DenseNet for Recognizing Driver’s Fine-grained Secondary Activities

“…Most of these sensing technologies are often focused on surrounding environmental perception and minimal work has focused on the human driver perspectives. It is suggested that mindful attention should be paid to driver behavior [11] since driver error is overwhelmingly to blame for the vast majority of accidents [35,1]. The driver's role could be replaced by the AVs but the vehicle also requires to deliver performance identical to that of a driver if it is to be trusted [21].…”