When Andy Schaudt called me, he was navigating around the Capital Beltway in Washington DC, not dressed as a car seat. But for 20 hours in August, he was.
Schaudt, project director of the Automated Vehicle Systems department at the Virginia Tech Transportation Institute, worked on an autonomous vehicle study that put a regular car—a Ford Transit Connect van—on the roads of Arlington County in Virginia with a driver that was camouflaged as a seat so that passersby would think it was autonomous. These drivers logged 50 hours on the road collecting data over approximately 1,800 miles, recording everything with high-definition cameras mounted around the vehicle.
The goal was to test how cars that drive themselves might communicate—using an array of lights and signals—with pedestrians and other drivers, and how people might react to a driverless car.
The study, which ran from August 1 to 31, might have gone largely unnoticed until Ford’s announcement of the details this week—if not for NBC Washington reporter Adam Tuss knocking on the van window and seeing nothing but a pair of hands and legs sticking out from the driver’s seat.
Now, we know the full story behind the dude who would not pull over, for science.
Why did you decide that disguising yourself as a seat would be the best way to do these tests?
We’re very experienced in doing real world data collection. One thing we pioneered was naturalistic driving research, which is outfitting the vehicle with cameras and sensors and letting people drive their vehicles for one to two years, and [collecting] all the data to truly understand what’s leading to crashes.
We wanted to look at the ability to communicate the intention of the vehicle. To make it look like it’s actually highly automated, we thought about just concealing the driver completely, or having the driver in the backseat driving the car. But what we decided to go with was, using psychology and misdirection in an attempt to conceal the driver right in the driver’s seat.
There’s been other efforts to do that. I guess you’d call them “ghost seats,” for research purposes and some for pranks, I’m sure you’ve seen those on YouTube. But other studies that have done this did them for short increments, collecting in a certain area. We needed to conceal the driver—actually, a lot of different drivers of different sizes, who were gonna drive this a lot. We needed to be a little bit more advanced in the design of our suit, so it can fit people of different heights and widths, and have good visibility.
It’s just [assembled] with magnets. I can lean forward and it will all pop off, so I can drive the vehicle safely in some kind of conflict. When somebody’s looking in from the outside, the windows aren’t tinted, so they can confirm their suspicions: Wow, there really is nobody in that car.
You can see our legs and arms if you got up close to the window, but it worked well.
What were some of the reactions people had while you were in the seat? Did people try to get your attention while you were in the car?
No, actually. There’s been some research out on the West Coast talking about people playing with the vehicle, jumping out in front of it and whatnot… We didn’t experience anything like that.
That reporter [Adam Tuss at NBC Washington] tracked us down and was very intent on figuring out what was going on. But that was really the only instance of that. For the most part, people were just going about their day. When they observed this vehicle with flashing lights and confirmed that nobody was in there, they would just kind of stop and look, or pull out their phone to try to get a video of it. But there wasn’t any kind of concerning behavior.
That may very well be because we don’t have LIDAR [Light Detection and Ranging] equipment all over the top of this vehicle, like the other automated vehicles. This is more of a simulation of what a car would look like 10 years from now once all the technology and sensors have been shrunken down.
How long were the shifts as The Seat?
Usually just a couple hours and then switch off. We tried to keep it around more active times in Arlington, so we’d do the morning shift when people are heading to the metro and going to work, then take a break, and around lunch time head back out and get lunch time crowd, when they’re out trying to relax and get some food.
We never did more than three hours per shift, usually a little less than that.
That face mask headrest thing that you put on. What did it feel like, looking out of that?
We have two different ones: One for nighttime that uses some clear plastic and fabric over it. And then the daytime one that uses plastic with a little bit more of a sheen to it, like a sunglasses lens. We put this fabric over it to match the black upholstery, so when you’re outside looking in, it looks like it’s really dark. But when you put your head inside and look through it, you can see very well.
I know it’s early and you just wrapped up this experiment, but just from first impressions, what do you feel like you learned?
None of us observed anything that would make us concerned for safety, so with regards to lighting and signalling and communicating, nobody was a deer-in-headlights… so that’s good. That’s something we feel at least confident about, that there was no harm or safety concerns.
Other than that, for the most part people are going about their days, they’re busy, and when they do see [the apparently driverless car] they’re moving across the sidewalk and maybe will stop afterwards. But there’s no people screaming and angry about anything.
There’s been some concern about automated vehicles in the environment because people haven’t experienced them, so I think this may be valuable to look at later, to see how people react when they see us. Automated vehicles may very well be accepted in these urban environments.
Living in a city, you see a lot of weird stuff in general. Maybe this is just another weird thing. Instead of panicking, you get your phone out and tweet it.
People are definitely engaged with their phones. Pedestrian and bicycle fatalities are on the rise, and that’s concerning, so anything we can do to help communicate with pedestrians who are most likely engaged with their phones and whatnot—that may be one of the influential factors in this.
I think the data we collected could potentially be helpful for other people as well. Ford has been very generous and public about this, and we’re happy to share the data and the results—if you feel like you can get some more data out of what we did, we’re happy to share.
This interview has been edited for length and clarity.