This is the first part of a new series on how bicyclists, pedestrians, and autonomous vehicles can cohabitate.
When I’m pedaling down North Pershing Drive in Arlington, Va., dodging cars or car doors or trash and delivery trucks, Mobility Mama wonders how her kids ever make it anywhere incident free.
“We never ride on Pershing, we cross it and head through the neighborhood. Mobility Mama, what were you thinking?”
“I was thinking about how much safer things would be if there were no drivers,” I offer lamely.
“Whatever. Just keep your mind on the road you are on and keep off streets where the only thing predictable is that no one will see you, no one will slow down, and that you are the loser in the law of great gross tonnage.”
Isn’t it delightful to have such a concise, pointed set of safety encomiums thrown back at you when you are just trying to puzzle out an important policy issue?
To the air, I ask, “Seriously, would things be safer if we had autonomous or driverless vehicles? At least the AV would stop if I biked in front of it.”
(I continued with my transportation-wonk musings, but my kids’ attention was in the proverbial rearview mirror and, if they could, they would have taken away the keys to my bike to prevent me from riding down a dangerous roadway!) Regardless, we were onto something.
As technology companies and automakers accelerate their testing and move to deployment, integrating AVs into environments where we want to encourage bicyclists (and pedestrians) is one of the toughest problems facing communities.
Dense urban communities are often top of mind when we think of safe biking and walking. But suburban and rural places are also crucial to thinking about interactions between AVs and bicyclists and pedestrians.
If a cyclist, for example, can stop a car or shuttle on a whim, the mobility and safety benefits from AVs may be reduced or erased. How can we keep bicyclists safe and reap the benefits of AVs?
Why can’t AVs just “see” bicyclists?
“Bicycles are probably the most difficult detection problem that autonomous vehicle systems face,” said University of California Berkeley research engineer Steven Shladover in IEEE Spectrum. Bikes are much smaller than cars and come in a wider ranges of sizes, colors, and configurations. Some bring the rider over the handle bars or encourage an upright position, and still others are recumbent or have a tag-along affixed to them. Some have panniers, heads-up GPS displays, straw baskets, flags or a wagon or kid cart on the back.
My colleagues at George Mason – in partnership with robotic-taxi developer Zoox – have been working on something called the Deep3DBox algorithm. As noted in the IEEE Spectrum article, they’re finding that their software is 88 percent accurate in telling whether a car is facing forward or backward and only 59 percent accurate in orienting a bike. That current level of accuracy for bikes is too low to rely on in current practice, but it’s likely to get better as technologists are able to train AV robots using large amounts of data about bicycles on the streets collected by firms like Mobileye, which was recently acquired by Intel.
As this work proceeds, different manufacturers have made some strides in connecting sensing technology to collision-avoidance such as automatic-braking technology. The missing link is figuring out how to predict the likely movement of the bicycle.
While automakers and researchers work to solve challenges of making AV vehicles safe to operate in environments where bikes and pedestrians are present, there is promise in safer road-design solutions, like those recommended within a new report from the Governor’s Highway Safety Association. Some of those include physically separated bike lanes and redesigned intersections.
Part 2 of this series will examine what communities can do to protect bicyclists even if driverless vehicles can’t yet fully see them. Part 3 will look at how AV-related technology can be used to make pedestrians and bikers safer well before we see lots of autonomous vehicles on the road.
Photo by the League of American Bicyclists.