Mechanisms of Visual Information Acquisition in Driving

点击翻译按钮获取中文摘要

Imagine you are driving down the street, and someone steps into the road ahead of you without looking to see if you have seen them - how quickly can you understand that the situation has changed? What do you need to know to stop in time, and how do you learn what you need to know about the situation in time to avoid a crash? What does your car need to know about what you know, and how can knowing all of this make the road safer for everyone? Right now, there is a push for more self-driving features in cars, but doing this safely means that we have to understand the person in the driver's seat more completely than we do. If, for example, your car is driving you to work, but encounters a situation it cannot handle and asks you to take over, what does it take for you to understand what is going on, much less to avoid the moose that just stepped into the road? Fundamentally, this is a question of human visual perception more than it is a question of engineering: how do drivers understand dynamic, real-world situations, like the road ahead of them? Drivers do not have time to look at everything in the world before events overtake them, so how do they get the information they need when they need it, how do they learn to do this, and how do their abilities change as they age? We can look at a photo of a real-world scene and understand it in the blink of an eye, using information from across the entire image, not just where we are looking, but what do we do in real situations when everything is moving? How do we do this on the road, where objects move around us and we move through the world? Research on driver behavior talks about the concept of Situation Awareness, our awareness of our environment and what might happen, but does not ask how we get the information we need to develop this awareness. This proposal builds on my work on Information Acquisition, where I hypothesize that much of what we need to know about the world we acquire with peripheral vision, that is, from the rest of our field of view away from where we are looking now, and that we look at specific things or locations in the world only when we need more detail about something specific. In this proposal, I ask: 1)When do drivers need to look at particular things or locations when driving, and when does the rest of their visual field give them what they need? 2)How much information about our environment do drivers need and what can they do without? 3)How do drivers learn to balance our need for information with the tools they have to acquire it, and how does this change as we learn and age? Answers to these questions are critical for designing next-generation vehicles that account for what drivers can and cannot perceive, and developing this understanding will help us better understand how we perceive the world, and help make the road safer for everyone.