It’s now apparently legal to have self-driving cars in California and Nevada, and this should spread across the country rapidly. One industry report predicts we’ll have them by 2019. For the purposes of this article, let’s assume that the costs will come down slowly but surely and adoption will grow quickly. Let’s jump all the way to the end point, where self-driving technology is safe, reliable, and mandatory (yes, mandatory), just like seat belts, air bags, and so forth.
Last time I wrote about robocars, I focused on the computer security threats (and the risk that hackers will steal your car remotely). This time I want to focus on the regulatory and implementation issues. Consider the case of the humble four-way stop. Today, you’re legally required to stop, even if there’s nobody there. Your robocar, however, could send out a message. “Anybody approaching this intersection?” If not, you blow on through. (Saving time and energy!) Likewise, if every car can compute exactly when it will arrive at the intersection, that means they can negotiate with one another. Maybe you speed up a little and I slow down a little and we nicely miss each other by a few inches. Sounds great, right? Extrapolate a little bit more, and traffic lights become completely unnecessary. Instead, you’ll have cars flying through the intersection, seemingly at random, but always managing not to hit each other. Even if a car experiences a tire failure or other catastrophic event, it can communicate that to everybody nearby, and they’ll respond quickly and safely. (But see my caveats below.)
Now consider that all the Google robocars have a big red button next to the steering wheel that forces the computer to disengage and return the car to your manual control. If you freaked out in one of these busy intersections and hit the big red button, everybody else’s scheduled entry to the intersection is now at risk. Nobody can predict what you’ll do next. Consequently, you could be liable for the damage caused by taking manual control of your car!
Of course, in the future, we’ll still have pedestrians and we’ll still have bicycles, roller skates, pets, and so forth. While a car can negotiate a very specific plan to go through the intersection, pedestrians and bicyclists will almost certainly still be subject to their current constraints. This leads to an interesting question of how robocars and pedestrians will relate. We could retain the current press-to-walk buttons and walk/stand signals. We might instead put fancier sensors that detect your pedestrian presence and telegraph it to every car approaching the intersection, forcing them to slow down and accommodate you, even if you’re jaywalking in the middle of the street. We might also require pedestrians to carry beacons that telegraph their location (maybe building this into their super-duper smartphones) and use those phones to tell them “please wait 30 seconds and then traffic will open up for you.” While this will work great for most cases, there will always be exceptions. For example, unless global warming kills off all wildlife, we’ll still have deer and other critters with which to contend.
Recall William Gibson’s famous quote, “The future is already here — it’s just not very evenly distributed.” That tells us a good deal about how we’ll ultimately solve these problems. There will be high-traffic intersections and roads that will be heavily instrumented: Total Traffic Awareness. Likewise, there will be lower-traffic intersections and rural areas where the cost/benefit of modern instrumentation won’t justify it. When our robocars have more information, they’ll be able to drive more aggressively. Without this information, they will necessarily revert to our current, more conservative traffic behaviors.
But what about your award-winning, meticulously restored V8 muscle car? If you want to use it on public roads, and it doesn’t have a robo-drive controller, you may be restricted to only driving off hours. You will almost certainly be required to have a transponder to warn all the other cars, and you’ll pay a lot of money for the privilege of driving it, since you’re slowing everybody else down. And, of course, you’ll be unhappy at the lack of parking spots, since the robocars just drop off their occupants and head off to a remote garage somewhere. Maybe I’m wrong, but you probably won’t want to drive your classic car except on special occasions.
Caveats: One of the many obvious gains to be had from robocars is that they can form freeway “trains” with minimal spacing between them. This improves road utilization and saves energy, since the lead car in the train is breaking the wind for everybody else. But what if something goes wrong, mechanically, with the lead car? Time will pass before the pack has figured this out and has begun to take countermeasures. That might not be enough time, particularly as you pack the cars in tighter. The same problem would occur in instrumented traffic intersections and anywhere else where cars are negotiating over their future locations. What’s the solution? More information and better predictions. Say the car in front reports it has low tire pressure on one wheel and the tire is nearing the end of its rated treadwear; it might still be perfectly drivable, but the cars behind will compute an increased probability of a tire failure and will give the lead car a little bit of extra room, just in case. They might even compute which way the front car is likely to spin and preemptively stay to the other side. Better safe than sorry, right?
For this to work, we’ll need two things: correctness and trust. We need all of these sensors and car-to-car messaging protocols to work correctly, but we also need some assurance that nobody’s cheating. If any car is making stuff up, they could cause all sorts of mayhem for the other cars in the neighborhood. Computer security academics have been working on this problem for almost a decade now (see, for example, this research group in Switzerland), but the legal side of the equation is pretty interesting. Car makers will bend over backward to avoid having liability for crashes. In order to do that, we can predict that they will have standardized, government-approved software (“don’t blame us; you were running the standard package, same as everybody else!”), along with tamper-resistant mechanisms to keep you from monkeying with that software. Maybe automotive tinkering will still be allowed, but expect it to be treated the same as that V8 muscle car. You tweak your car, then you’re restricted in when and where you can drive it. Yeah, this sounds a bit like a dystopian future, but it mostly indicates the transition from cars as a romantic possession to cars as a boring utility to get you where you need to be.
Related reading: science fiction author and computer scientist Vernor Vinge has a great book called Rainbows End. He describes a near-future to our own that’s full of interesting gadgets. His vision of what cars might become is pretty close to what I’ve been talking about here.