We hear lots of stories about autonomous cars hitting the road and the kinds of tests automakers are doing to make these cars ready for the market. But there’s an important part of the conversation we rarely hear about: how will the safety of autonomous cars incorporate data on the human attention span into its plans, and what role would distracted driving play in these cars?
I recently talked with Bryan Reimer, a research scientist at the Massachusetts Institute of Technology, about this. He leads a number of research efforts looking into the interaction between autonomous car safety and the human attention span. Bryan is associate director of the New England University Transportation, and he founded three consortiums dedicated to researching the relationship between autonomous vehicles and driver attention spans, including the Advanced Vehicle Technology consortium.
I began our conversation by talking with him about how he entered this field of research and automation’s impact on autonomous car technology.
Automation in driving has been creeping up on us for years with things like antilock brakes, cruise control and the automatic transmission; now take them for granted. We don’t realize how, little by little over time, this automation creep has disengaged us from the task of driving. The automatic transmission freed one of our hands and created an “attentional void” we strive to fill somehow by grabbing our smart phones or a sandwich. With a dramatic increase in driving automation that’s soon to come, the void will become even bigger. We will become even more disengaged. But how does our robotic car get us reengaged when it is in over its head an AI-powered head? That’s the big question.
I next talked with Bryan about the handoff from autonomous cars to the driver and how autonomous cars can detect a human’s attention or inattention.
ED BERNARDON: So when a driver’s attention is not required to drive, they’re going to wander off and potentially sleep or daydream, talk, play cards or whatever it might be. You talk a lot about the handoff problem of how do you design a system so that when the automation can’t handle the situation, it tries to give control back to the human driver to keep them safe?
How do you design an automated system, or make an autonomous or even driver assist vehicle, safe when this handoff must occur?
BRYAN REIMER: That’s the million dollar question these days. It’s clear that it requires an understanding of the extended situation to in essence provide a window of time for the transition to occur. If the sensing system fails instantly, then handing over attentional or management responsibilities to the driver instantly would be quite problematic, so we need to project driver awareness of the operating conditions forward to buy time.
We need to manage effectively the driver’s attention inside the vehicle so that we know, at a moment-to-moment basis to project forward, how long it may take them to regain control. We need to fuse those two decision logics together to manage the situation. Now if we believe the driver is not capable of taking control in the required period of time, then we need to have a safe harbor plan that is not stopping in the middle of an interstate.
This whole topic is an area of incredibly interesting studies that we focus on through the Advanced Vehicle Technology consortium where we’re advancing the understanding of driver workload metrics for use in the overall understanding of driver awareness and attention, and to develop systems that help manage attention in this increasingly complex operating environment.
ED BERNARDON: Is any thought being given to keeping the driver attentive? Could you do something as simple as poking them, flashing a bright light or blasting “Hey, you’re dozing off!” Can you even detect that a driver is inattentive?
BRYAN REIMER: Absolutely, and a key part of our research is to providing avenues for system support for attention management. How do you support attention longitudinally so we can help manage drivers?
That’s got to be brought into a cohesive framework for attention management — it can’t be looked at as an outside force. It really has to be managed into the equation. It’s an area of study where we’ve put an incredible amount of investment involving OEM, suppliers and large technology companies.
ED BERNARDON: Automated driving systems seem to be evolving in different ways. There’s a top down approach like Google’s where you automate everything, or bottoms up, where you take smaller steps beginning with ABS [anti-lock brake systems] and automatic transmissions, then cruise control, to eventually what you get in the Tesla Model S today.
How do the challenges that these alternative technology development approaches pose for human driver integration differ?
BRYAN REIMER: The challenge is – and this is I think one of the reasons that Google went to the top-down approach — they figured that it’s a technology challenge, there is a human interaction piece, but it’s not as complex.
The bottom-up — the human is probably the more complex piece and the technology is probably the simpler piece. Again both are quite complex in nature by themselves but the thinking about these is what is the percentage contributions.
This concludes the second part of my conversation with Bryan Reimer. In part three, we discuss the current state of trust around systems in autonomous cars, as well as the potential training drivers will have to take whenever autonomous cars receive significant updates.
About the author
Edward Bernardon is vice president of strategic automotive initiatives for the Specialized Engineering Software business segment of Siemens PLM Software, a business unit of the Siemens Industry Automation Division. Bernardon joined the company when Siemens acquired Vistagy, Inc. in December, 2011. During his 17 year tenure with Vistagy, Bernardon assumed the roles of vice president of sales, and later business development for all specialized engineering software products. Prior to Vistagy, Bernardon directed the Automation and Design Technology Group at the Charles Stark Draper Laboratory, formerly the Massachusetts Institute of Technology (MIT) Instrumentation Laboratory, which developed new manufacturing processes, automated equipment and complementary design software tools. Bernardon received an engineering degree in mechanical engineering from Purdue University, and later received an M.S. from the Massachusetts Institute of Technology and an MBA from Butler University. He also holds numerous patents in the area of automated manufacturing systems, robotics and laser technologies.