Google builds its own self-driving car, sans steering wheel

Google's testing of self-driving cars has always generated more questions than answers. While it demonstrated the viability of such technology on public roads, and spurred major automakers to vow that their own systems would be in public hands by the end of the decade, there's never been a clear answer as to where Google planned to take its own technology.

Today the answer arrived in the form of the illustration above: Google will build it's own self-driving cars  — ones that test the technology to such a degree that they lack steering wheels.

After several years of testing Toyota Priuses and Lexus RX 300s outfitted with sensors and Google software, Google's director of autonomous vehicles Chris Urmson revealed the plan Tuesday night in a blog post, saying the company was aiming to build about 100 prototype vehicles, equipped with two seats, a little room for passengers and a few buttons for inputting directions — and that's about it.

"They won’t have a steering wheel, accelerator pedal, or brake pedal… because they don’t need them," Urmson said in a blog post. "Our software and sensors do all the work. The vehicles will be very basic—we want to learn from them and adapt them as quickly as possible—but they will take you where you want to go at the push of a button."

Google's electric-powered cars will have a 100-mile range and be limited to 25 mph — putting them under federal rules for "neighborhood" vehicles that require seat belts, headlamps and wipers, but not a full suite of air bags and other safety systems that would require far more complex automotive engineering. (Those same rules require rear-view mirrors, even though the Google passengers can't do much about what they show.) The front of the car and the windshield were also designed to minimize harm in any impacts with pedestrians or cyclists.

Urmson said Google would put the cars in employee hands on private roads in a few months, with the first drive on public roads by the end of the year. Google then hopes once it has proven the safety of its software to recruit testers from outside the company within a couple of years. In order to comply with California law, the cars will have to have some form of "manual controls" for emergencies, which Urmson said Google was still developing.

The company has racked up 700,000 miles of self-driving tests in its modified vehicles, and Urmson said the idea for its own car came from behavior of its drivers, who would so trust the system when it was on the road that they would be "too distracted." That's where the idea of removing the steering wheels and pedals was born.

"We have this problem of de-bugging the human," Urmson said. "Given that our overarching goal was improving society, we decided we can either spend time trying to work on the human factors problem, or take a step forward on the fully self-driving capability."

The search-engine giant's move into building its own vehicles steps up the competition with traditional automakers, who have preferred to explore designing their own technology rather than partnering with Google so far. All major automakers have demonstrated some level of self-aware vehicles; the new Mercedes S-Class models can control themselves for a few seconds at a time, and Nissan has vowed to build a self-driving vehicle for everday buyers by the end of the decade.

While Google and other backers of the technology tout the potential safety benefits of leaving driving duties to software, many legal and moral questions have yet to be answered. In its current form, Google's system can't drive roads that have never been previously mapped; the laser imaging system used for scanning obstacles today can't distinguish the density of objects (a cardboard box looks like a concrete pylon.) Many automakers have questioned whether their systems would faced increased legal liabilities in widespread use. Even Tesla co-founder Elon Musk has said it was more likely that future cars would still require some level of human input akin to how airline pilots monitor autopilot systems.

Urmson said Google did not have a detailed business plan for its technology yet, but that the benefits could extend well beyond safety, to functions like shared-car services. Cars today sit idle for most of the time, and Urmson called them "kind of a poor capital investment" that raises the cost of personal transit beyond many people's means.

"If you could call one of these vehicles, it could take you where you want to go and then go off somewhere else," he said. "The cost of transportation could be much lower."

Whatever questions self-driving vehicles pose, it's clear that Google intends to answer them in the real world rather than wait for the auto industry to take away the steering wheel. And if the safety benefits can be demonstrated, especially for intractable problems like drunk drivers, the Google car could be the start of a revolution in how we drive — or not.