Self-Driving Cars Should Accommodate People, Not the Other Way Around

A basic principle should be that autonomous vehicles have to rely on their own sensors, not devices connected to pedestrians and cyclists, to operate safely.

If the sensors on autonomous vehicles aren't safe enough on their own, the vehicles aren't good enough for crowded city streets. Photo: Richard Masoner/Flickr
If the sensors on autonomous vehicles aren't safe enough on their own, the vehicles aren't good enough for crowded city streets. Photo: Richard Masoner/Flickr

A segment that aired yesterday on NPR’s Morning Edition comes off as posing a reasonable solution to the technical challenges of making self-driving cars safe around cyclists. But beneath the techno-optimism, there’s a dystopian vision in which being wired up is the only way for people walking or biking to avoid traffic dangers on the street.

The story begins innocently enough. Reporter Margaret J. Krauss, from WESA in Pittsburgh, talks with an autonomous vehicle researcher from Waymo about how the technology detects cyclists. She also learns from local bike advocates that many riders feel more comfortable around self-driving cars, because they aren’t subject to road rage.

But things take a turn when Krauss interviews Anthony Rowe, an associate engineering professor at Carnegie Mellon University, who showed her a bike “loaded with gear” to send information to nearby autonomous vehicles:

“What we’re trying to do is put as much instrumentation on a bike as we can to see if we can predict how it’s going to move in the future, so that it could, for example, signal a collision warning system on a car,” he says…

“I would not be happy if I had to ride this every day,” says Rowe, hopping off the bike. “But hopefully when all of this stuff just gets embedded in a cellphone on the front, then it should be no problem.”

Krauss then concludes her piece: “Rowe thinks self-driving cars will make the future a lot safer for cyclists and pedestrians,” she says. “But while humans remain the primary pilots, he thinks a little help from bikes could compensate for their weaknesses.”

Oh, dear. Where to begin?

The interest in attaching sensors to cyclists stems from the inability of current automated vehicle technology to safely interact with bike riders, who take up less space and move less predictably than people driving cars. One AV algorithm, for example, can spot just 74 percent of cyclists and determine the orientation of just 59 percent, far below the rates for cars and trucks, reports IEEE Spectrum.

Wiring up bikes and pedestrians makes sense in a testing environment, when self-driving car technology is being refined. But at no point after they’re operational on public streets should they require sensors on people walking or biking. A basic principle should be that AVs have to get by with their own sensors.

Designing a future in which people outside of cars can only expect to be safe on the street if they launch an app or have a high-tech bicycle is a recipe for disaster.

You think victim-blaming is bad now? Making everyone walk or bike with a “don’t hit me” device would further penalize the most vulnerable. Old? Poor? Lost your phone? It’s your own fault for walking into the crosshairs of robot cars.

Other reporters have explored the pitfalls of bike- and pedestrian-mounted detection devices. Last month in the Guardian, Laura Laker spoke with academics and experts about how such initiatives put cars before people. “Cities have some urgent questions to answer,” she wrote. “Failure to address the issues raised could see us sleepwalking back into the problems of the 1960s and 70s, where cities became thoroughfares for traffic first… and places for people second.”

Bike advocates have also raised concerns about sensor-mania. “Should we be forcing cyclists and pedestrians to be equipped with detection devices when they should be prioritised in urban areas?” the European Cyclists’ Federation asks in a policy paper on autonomous vehicles. “Should we also equip children walking to school? Do we want to live in urban areas where we have to remember keys, wallet, phone and protective [Connected Intelligent Transport Systems] device to stop cars running into us?”

“Of course, this sort of debate brings us back to the fact that we know how to make cyclists and pedestrians safe in urban areas: reduce speed, shifting cars away, and separate infrastructure,” ECF says. “Why over-engineer the solution to a known problem?”

  • Jeffrey Baker

    I think Waymo would push strongly against this, since their cars are able to safely navigate around pedestrians and bicyclists. It is the other companies, who are all way behind, who want to take a shortcut here.

  • Larry Littlefield

    “Designing a future in which people outside of cars can only expect to be safe on the street if they launch an app or have a high-tech bicycle is a recipe for disaster.”

    I agree. They can’t rely on it. What if a child runs into the street after a ball? No sensor, tough luck?

    On the other hand, there may be a way for those to are working on self driving cars to have to sensors for those walking and biking incentivized.

    By compiling the amount of confirmed walking and biking, for forwarding on to health insurers who would the pay for active transportation. The way they are starting to pay for Citibike, for which where are records of how much it is used by each user.

    https://www.villagevoice.com/2017/07/10/hop-on-a-citi-bike-doctors-orders/

    Basically, since Generation Greed’s economic orgy has left everyone broke, any worthwhile innovation has to save money not cost it. Perhaps AVs and active transportation apps could save money by sucking it out of the auto insurance and health insurance industries.

  • Vooch

    level 4 ?

    or

    level 5?

  • Jeffrey Baker

    I believe Waymo claims to be targeting Level 4. As I recall, the difference between 4 and 5 is whether or not the car can operate independently in a whiteout blizzard. I don’t think the difference is really germane to pedestrian safety.

    The other companies are all at Level -1. Their systems are worse than a meat-driven car. GM reported that their drivers had to intervene once every 100 miles. Waymo reports once per 5000 miles.

  • bettybarcode

    “Self-Driving Cars Should Accommodate People, Not the Other Way Around.” The headline says it all. Driving is a privilege; walking is a right. Not the other way around.

  • Stuart

    How do you get from

    “But while humans remain the primary pilots, he thinks a little help from bikes could compensate for their weaknesses.”

    to

    The interest in attaching sensors to cyclists stems from the inability of current automated vehicle technology to safely interact with bike riders

    ?

    It’s still a completely backwards way of thinking about preventing collisions, but the phrasing is very clear that this is about communicating with the collision detection systems in current/near-future cars to compensate for bad human driving, and has little if anything to do with self-driving cars.

  • Tom

    Reports of collision with level 1-3 AVs will be quick to point out that “the bicyclist was not using a transmitter”, while the data showing who was in control of the car (human or computer) at the time of collision will be deemed private and not accessible by investigators. The reckless driver who was in control at the time only needs to blame the level 3 car, and the car maker blame the bicyclist. This will be followed by a ‘safety’ campaign by the local police to ticket bicyclists who don’t have working transmitters…in low income neighborhoods of course.

  • reasonableexplanation

    What about riding? In time being in a self driving car will be no different than taking the train, just smaller. Privilege or right?

  • Al Mundy

    Cute cliche but the distinction seems flawed.

    A driving license is a right as long as you pass the tests. With a DL you can drive anywhere it is legal to do so.

    Walking can be a privilege in many circumstances. Pedestrians have the right of way in some locations, have fewer rights in other locations, and no rights at all in some places.

  • Joe R.

    It seems to me this is an admission by autonomous vehicle designers that they are currently unable to design algorithms/sensors which can detect pedestrians and cyclists with, say, 99.9999% probability. Attaching sensors is one possible work around to the problem but there are two others:

    1) Just keep improving the methods as better sensors and AI allows you to. Remember it’ll probably be at least a decade before most vehicles are autonomous, probably a generation before they all are. Seeing how computing has progressed in similar time spans, I’m actually optimistic we’ll solve the detection problem in that time frame.

    2) If #1 fails then restrict vehicles primarily to limited access highways in urban areas. If they go on local streets, then they’ll have to proceed at a pace where there is very little chance of injury if they collide with a pedestrian or cyclist. External vehicle design, such as adding cushions which inflate during urban driving, could make “safe” speeds somewhat higher. If we get to the point where a vehicle can hit a child at 20 mph with little chance of injury then we have a travel speed on urban streets which is a reasonable compromise of efficiency and safety.

    Remember once all vehicles are autonomous there is no longer any need for traffic controls at intersections. The rationale here is pedestrians/cyclists won’t obey them, so vehicles will need to be programmed to avoid them anyway. That being the case, you might as well just program the vehicles to give priority to cyclists and pedestrians at intersections all the time. This would be a tremendous improvement over what we have now.

  • Stuart

    It seems to me this is an admission by autonomous vehicle designers that they are currently unable to design algorithms/sensors which can detect pedestrians and cyclists

    If you click through and read/listen, you’ll see that the only person in the story who is actually an AV designer said exactly the opposite, and that the headline is just clickbait:

    “You want to predict what they’re going to do next,” he says. That’s something Waymo’s vehicles have learned to do over 1 billion simulated miles of testing and 3 million miles of driving themselves.

    It’s a skill that allows the cars to “fend for themselves,” an idea Fairfield says has long motivated his work.

    The 74%/59% number quoted in this post is about an algorithm that evaluates 2D images without any context. Would you like to guess how many AVs are going to hit the road using that approach? Notice that if you click the link it’s an algorithm from a company you’ve probably never heard of, rather than any of the companies that are actually competing in the AV space?

    It’s disappointing to see scaremongering about AV safety on StreetsBlog given that the best of the AVs today are clearly already better than the vast majority of drivers on the road, and are likely to continue to get better while human drivers stay the same on average (or get worse, as smartphone use while driving becomes increasingly normalized).

  • Isaac B

    Here’s a link to the driver manual in my state. You won’t find “driving rights” in there. You will find “privileges” a lot. It was even a question on my permit exam.

    Sure, you don’t have the “right” to run out in to traffic, walk along a freeway or cross against a light. But by and large, by law, walking is a right. Driving is not.

    https://dmv.ny.gov/brochure/mv21.pdf

  • Al Mundy

    I don’t find “pedestrian rights” anywhere either. It’s a bogus comparison.

    All that matters is that there are places where pedestrians have priority and places where vehicles have priority. The rest is pedantry

  • And Waymo announcements are completely free of hype.

  • Larry Littlefield

    Could be good, could be bad, depending on the rules. But the key is speed.

    If AVs turned the corner at no more than 5 mph, they could stop on a dime if a pedestrian suddenly appeared. Whether the AV’s fault or the pedestrian’s fault.

    Similarly, if AVs slowed approaching an intersection, they would be less likely to collide with people or other vehicles in that intersection, no matter who had the ROW.

    Higher speeds mean less of a chance to react and prevent a crash in the event of a mistake. Whether by the AV or a human. But an AV need never be distracted, and never speed.

  • Keninoz_1

    Hopefully these issues will be fully addressed as part of the R&D of autonomous vehicles before they are allowed on the roads. If not, they should either not be allowed on the roads or restricted to places where cyclists don’t ride, places like freeways.

  • nlpnt

    Under no circumstances would *a* don’t-hit-me device be sufficient, either. You’d have to carry both a Google/Waymo and an Uber one, at minimum, due to incompatible standards. Three if Tesla goes their own way.

    And you think it’d be an app in the phone you have anyway? Well, it’d be one (two, three…) that would use way too much memory and guzzle battery life since the developers wouldn’t have any incentive to make it otherwise…

    Some time ago Jason Torchinsky of Jalopnik suggested that there be some sort of external indicator showing that a vehicle is operating autonomously – he suggested a purple light (since that color isn’t otherwise used) showing front, rear and side. I’m beginning to think that’s a really good idea.

  • neroden

    Nope. A driver’s license is NOT a right, even if you pass a test. You can have it revoked for pretty damn near any reason whatsoever, and you can be prohibited from driving pretty damn near anywhere at any time.

    (Except on your own private property; *there* driving is a right, and you don’t even need a driver’s license.)

  • neroden

    Nope. There are pedestrian rights, enshrined in common law and judicial rulings dating back to the 10th century and before. The right to walk on the “public highway” (road) is an actual RIGHT. Congress had to pass a very specific exception to make the freeways possible. (The freeways are *not* public highways for common law purposes, legally!)

    You have a right to walk along the “public highway” everywhere else, period, end of story.

  • Andrew

    Smaller per vehicle but much larger per person. Space isn’t free.