The autonomous car movement has its origins in the need to make roads safer by freeing them of a common recklessness that regularly puts life and limb at risk. Basically, self-driving cars have been trying to make driving as inhuman as possible. But recent test results reveal that making them more human may be the way to go.
Aiming for Imperfection
Google’s autonomous car prototypes are very good at abiding by the rules; they come to a complete stop at stop signs, they maintain a safe distance from other vehicles at all times, and they never go above the speed limit. While this all sounds good, this degree of blind obedience is proving problematic; they’re a little too good at following the rules.
As the Wall Street Journal reported, Google cars hesitate more often than probably necessary, frequently stopping at seemingly unusual times. In one instance, a car quickly approaching from behind caused an unexpected application of the brakes. In another, a pedestrian approaching the corner of a sidewalk cause a complete halt in the middle of an intersection. Then there was stopping behind parked cars because the cars weren’t allowed to change lanes over a double-yellow.
Imagine driving behind one of those.
Google has come to realize that if their cars are really going to survive on the roads – and keep the rear-ending to a minimum – they’ll have to start bending the rules.
They are now on a mission to make its cars, well, more human; eager to cut corners, a little selfish, and less-than-Boy-Scout safe. Even for Google, this won’t be easy. Humans are much too complex to have their real-time decision-making mapped into an algorithm, especially when highly variable emotion and instinct comes into play. Overall, as a Google team lead told the WSJ, “You want to figure out when the right time is to relax things a little bit.”
A Little Background
Manufacturers have been toying with the idea of autonomous technology a lot longer than you’d think. Way back in 1925, Houdina Radio Control presented a prototype of what was basically the life-size equivalent of a remote-controlled car. The ‘Linrrican Wonder’, as it was called, was controlled by radio signals routed to its antennae by another car following closely behind it. Not the most efficient technology but a predecesor to modern autonomy nonetheless.
By the 1960s it was clear that self-driving cars needed to be able to do three things effectively: sense, process, and react. While existing technology could tackle the first and last of those to a respectable degree, it was the ability to process environmental information that posed a real challenge to the industry.
The first big breakthrough came in the 1980s through the work of German AI pioneer Ernst Dickmanns. Along with his team, Dickmanns refurbished a Mercedes-Benz van to make it possible to control its steering wheel, brakes, and throttle remotely. Software was written to convert input from cameras and other sensor-based feedback into commands to guide the navigation of the car. The 5-tonne Mercedes was taken hundreds of miles across Germany using just its autonomous technology to guide it.
The Defense Advanced Research Projects Administration (DARPA) has had a role to play in accelerating the development of autonomous technology. Its annual $1 million competition that started in 2004 laid the groundwork for making American outfits serious contenders in the race towards a commercially viable self-driving vehicle.
Now, smart vehicles innovation is heavily sponsored by corporate players as more and more people are sensing the lucrative potential of being a first-mover in the industry. Brands like Audi, Mercedes-Benz, and Toyota have been hard at work with their own branded technology. There are also companies like Cruise, NVidia and MobileEye, that are dedicating themselves to technology that will make any car smarter.
As it stands, autonomous cars are legal in the United States. But that’s just because they’re not illegal; there isn’t a set of rules catering to their particular category of transportation. While the general consensus is that the technology will get into high gear eventually, a robust legal framework guiding its implementation is needed before the rubber meets the road.
State like California, Florida, Nevada, and Washington DC have promulgated laws to oversee how self-driving vehicles are tested and made available to the public. However the lack of uniformity has been an impediment to manufacturers looking to make progress across regions. A few years back, the National Highway Traffic Safety Administration (NHTSA) stated that it had begun work on scripting the laws at a national level. Though this would help the cause of manufacturers, there is concern that regulators may not understand the technology nor have the vision to enable innovation in the field.
A Volume Game
Google’s travails with the humanization of autonomous technology will be a problem as long as the majority of us are stuck behind the wheel. When/if self-driving cars become common, there won’t be as much of a need to account for such human variability. To get there, however, we need the laws to catch up to the technology. Then, as with any classic roundabout, it will only be a matter of time before the technology catches up to the requirements.
Prateek Jose is a writer and engineering undergrad from India with an unhealthy obsession for obscure historical trivia. Conversations about absurdist fiction and the technological singularity make his day. He’s already uploading parts of his brain to servers by writing for websites such as this one.