Over the past few years, automakers and tech giants have been dumping billions of dollars into making the first true accident-free self-driving car. It’s been a dream for nearly as long as cars have existed: getting in your car, cranking up the tunes, and kicking up your feet as a self-correcting, self-driving computer swivels and swerves effortlessly through traffic better than any human ever could. but how close is that dream to a reality?
How Do They Work?
Driverless cars use an array of sensors, cameras, radars, real-time 3D maps, and gigabytes of specialized software to “see” the road in front of it, behind it, and around every corner. Operated by actuators attached to the driving column and pedals, self-driving cars take a constant stream of data coming from all corners of the vehicle and translate it into the motions of driving on freeways, city streets, and even suburban school zones.
By incorporating what the car can see into one coherent image of the road, self-driving vehicles are able to navigate almost any terrain in any weather condition, save for a few select instances where it still struggles to get a proper lay of the land (as we’ll get into later).
Right now, the two biggest players in the space of mapping and building the cars themselves are Google and Tesla. Each company maintains their own armada of retrofitted driver-based cars that were modified to become self-driving after stock, as well as a smaller selection of prototypes that were built from scratch on the factory line to be fully autonomous from day one. In fact, Google’s so confident in their self-driving models they’ve actually taken out the steering wheel and pedals altogether in the latest model, removing the driver’s ability to interfere with their program and leaving the rest up to the engineers.
Autonomous Vehicles Today
Not many people are aware of this, but we’ve already had semi-autonomous cars on our roads for years now. While they may not be picking us up from the store just yet, you can already get some self-driving features in premium packages of certain cars. They can parallel park themselves, press the brake if they sense an impending collision, or course correct the wheel if it detects the driver is drifting out of their lane on the highway without using a turn signal first.
These are automated systems that are integrated into some of the higher-end brands like Lexus, Mercedes-Benz, and BMW, designed to take some of the guesswork out of squeezing into a tight spot on the street or preventing that next fender bender from making you late to work. Ultimately they don’t amount to much more than an advanced form of cruise control, but they still use many of the same sensors you’d find in a fully autonomous vehicle (radar, distance-calculating lasers, etc.) to predict what other drivers are doing and react accordingly.
All of these features had to pass extremely vigorous safety testing and navigate through a gauntlet of bureaucratic loopholes before they were approved for consumer purchase, and it’s expected that the legal battle ahead for autonomous vehicles could take decades before they’re eventually approved to hit the streets. That said, both Google and Tesla have been testing their own vehicles on the freeways and around the suburbs of Silicon Valley and California for upwards of four years, often with no one even sitting in the driver’s seat (considered a sore point of contention by the state’s legislature).
In all of the testing that both companies have run in that time (1.2 million miles driven by Google’s fleet of 23 Lexus SUVs alone), driverless cars have proven themselves to not only be just as good as people at driving, but actually better than us in most cases. Their accident rate is lower than 0.2% (while easily-distracted humans average closer to 1.09%), and in the very few cases where the cars did get into an accident, it happened to be the fault of another person hitting them from the side or behind.
Thus far, the cars have shown they can run long distances without any problem (Google’s engineers have been taking regular trips to and from Lake Tahoe in the snow), and as long as it has updated mapping data of the area it plans to navigate, the only risk to anyone riding shotgun is if the road conditions suddenly change, say on a patch of black ice or during a hydroplane.
So Why Aren’t They Being Sold Yet?
Even with all these obvious benefits to adopting self-driving cars, there are still three major drawbacks that stand in Google’s way for total pavement domination: the lack of available mapping data, minor technical difficulties, and legal troubles.
The first issue is solvable, but it’s not going to be easy. When a self-driving car goes on a new road, the entire route it travels on needs to be 100% mapped by a normal car before the self-driving vehicle even knows what to do with itself. This means that for every street, dirt path and back alley way we might want to travel in the future, it will first have to be uploaded to a database, then mapped by Google, and uploaded to the hard drive of all driverless vehicles on the road.
This is obviously a monumental task in road-happy countries like the US, which means that before we can even push go on these cars, the Google Maps street team is going to have a lot of ground to cover both domestically and abroad.
Next, there’s the problem of the law, insurance companies, and deciding who’s going to be left on the hook if a car causes an accident. When we talk about automated driverless vehicles, it’s impossible to ignore where the moral and ethical responsibility of a person ends, and the actions of their car begins.
The question here basically boils down to some version of of “They haven’t killed anyone now; but what happens when they do?” It may not be today or tomorrow, but as time goes on we could be seeing weekly headlines about another accident that resulted in serious injury or death. Who’s held liable in that case? The company who made the car? The coder who programmed it? What about the person who was sitting in the drivers seat but didn’t react quick enough when the car took a wrong turn? When you hand the keys for a two-ton steel machine of death over to a robot, who ultimately takes responsibility in the 0.001% of cases when a bug or glitch ends the life of another person?
These are the scenarios that no one is too eager to address without running through a bit more data first. Though there haven’t been any injuries as the result of driverless cars being on the road (yet), the sample size of the activated fleet is so small compared to the number of human-powered vehicles on the road that it will only become more difficult to predict what the world might look like once that statistic starts to tip the other way.
Until we do enough peer-reviewed studies on the risk of bodily harm in wide-scale rollouts of driverless automobiles, the reality of seeing a self-driving car in every driveway on the block is still just a pipe dream wrapped in a haze of murky laws that haven’t even come close to being agreed upon yet.
Last, there are still some purely technical hurdles that the engineers on these projects will need to overcome before anyone is buying these things down at the local dealership. Namely, engineers need to solve what’s known as the “this or that problem”: when two separate driving choices present themselves at the same time, and the car needs to decide the best course of action. Say you’re driving along in a city, and the car takes a wrong turn right into a parade full of people thirty blocks long: should it attempt to reverse around a corner it can’t see, or wait 20 minutes until every float goes by? Which is considered the “safer” option when people start to surround the car from the back, too?
Driving is a lot of things, but at its core, it’s something that involves a lot of judgment and decisions–something humans still excel at. Until a machine can think at least as quickly as us about how to react to various new situations on the fly, they won’t be anywhere near as efficient as we are at getting through detours, roadblocks, or St. Patrick’s Day celebrations from the moment they pop up…but that doesn’t mean all hope is lost just yet.
RELATED: What is a Heads Up Display (HUD), and Should I Get One?
Like any new technology that’s hurled into the lexicon of public discourse, the adoption of fully autonomous vehicles will be a slow, but constant march forward into the unknown. Although it may be a while yet before the average driver is lucky enough to have a self-driving car parked in their garage, it’s predicted that the commercial transportation sector could begin adopting self-driving trucks and taxis en masse as early as next year.
Both long-haul truckers and taxi drivers could be the first to hit the unemployment line as a result of self-driving cars taking over the wheel, as multi-national conglomerates use their massive piles of spare cash to buy up the technology in bulk, while simultaneously using their roster of lawyers to push the necessary laws through state and federal courts to get it on the road. Tech startups like Uber have already begun placing orders for half a million autonomous taxis to be delivered from Tesla by 2020, while shipping companies like Freightliner just let their first fully-automated Inspiration 18-wheeler loose on the highways of Nevada back in May.
Autonomous cars could also reshape how we look at white collar work. Right now, people waste upwards of half of their entire day just commuting to and from the office, all valuable time that could be better used catching up on reports, video conferencing, or even typing up that monthly expense sheet. Once cars are self-driving, all the time we waste paying attention to the road could instead be spent in a “mobile office” of sorts, where early tasks of the day can be folded right into the commute. This means more time spent with our families, enjoying our hobbies, and less road rage buildup that explodes when you least expect it.
When Do I Get to “Drive” One?
Don’t start blowing off the commute to work just yet, though, because it’s still going to be half a decade or more before any state or federal agency signs themselves up as the guinea pigs for true consumer distribution models. Yes, both Google and Tesla’s self-driving cars have impeccable driving records for the time being, and yes, they have millions of miles under their belt where the only accidents were the fault of some other human. Realistically the technology that makes these cars work in all conditions isn’t more than year or two from being 100% road ready…but people fear change, legislators doubly so.
Fight as they might against the inexorable march of progress however- just as the introduction of the first car flipped cities and horse stables around the country on their heads – self-driving cars will inevitably shake up more than a few major industries before the next century is over, and we’re all just going to have to get used to it once they do.
But they will also bring with them a new way to work while we’re on our way to work, give us more time to interact with family or friends on long road trips, and once fully adopted, create entire nations filled with safer and more accident-free roads. Driverless cars are everything that comes to mind when you think about “the future”, and they’re only a few skips and a jump away from completely revolutionizing the way we get ourselves around.
Image Credits: Tesla, Flickr, Wikimedia 1, 2, 3, PixGood, Freightliner