How (and How Well) Google’s Self-Driving Cars Get Around

Having redefined the way people search for and find information on the internet, Google — everyone’s favorite/most-reviled search giant — has now set its sights on revolutionizing driving.

Google's driverless car
Photo by Steve Jurvetson. Trimmed and retouched by Wikipedia user Mariordo.

We think the most important thing computers can do in the next decade is to drive cars.—Sergey Brin, cofounder of Google

Having redefined the way people search for and find information on the internet, Google — everyone’s favorite/most-reviled search giant — has now set its sights on revolutionizing driving.

That’s right: Google has put its immense resources and technological ingenuity to use — radically transforming how people get from point A to point B. How? By taking the driver out of the equation.

But just how do these driverless vehicles of the (not-so-distant) future do it?

How Google’s self-driving cars work

Think of the driverless car’s systems as physical embodiments of the “spiders” Google’s search engine uses to map the digital world. These spiders crawl all over the web, indexing, cataloging, analyzing, and ranking information. Then they serve up search results for your browsing pleasure. The self-driving car “crawls” the real world, but makes its own decisions about the right path to take.

In the real world (in which an autonomous vehicle needs to get around), Google performs 2 information-gathering missions. The first is done by a good ol’ fashioned human being, who travels the car’s intended route, gathering information about any recent changes, such as road condition, new markers, etc.

That information is then uploaded to Google’s servers. Next, a backup driver and software engineer hop in the robo-car and load the previously gathered route data into the car’s computer system.

As the car sets off, a powerful laser array on top of the car, several smaller radar arrays, and a standard GPS start gathering new information about the live environment rolling by. The lasers map the car’s immediate surroundings out to a 50-foot radius. Using the lasers, the car can differentiate between other cars, pedestrians, cyclists, and small and large stationary objects (including death-defying squirrels, hopefully).

Meanwhile, the radar scans for fast-moving objects beyond the laser’s range. And a third sensor system, a camera mounted on the front of the car, gathers the same data the human eye would: traffic lights, road signs, etc.

Google’s computers synthesize all this data in real time, giving the car’s artificial intelligence the info it needs to make the right decisions. Decisions that include tricky choices, such as when to take the initiative if hesitant drivers don’t act on their right-of-way. It turns out the intelligent auto isn’t so robotically kind that it can’t opt for a little aggressive driving when the situation warrants it.

And should the human behind the wheel decide to take over, the car reverts to manual control with a tap on the brake pad, a turn of the wheel, or the press of a button.

The driverless cars’ track record so far

To date, Google’s mini-fleet of self-driving cars has logged over 250,000 miles. Almost all in California, where laws covering the artificially intelligent vehicles don’t yet exist.

But how’s their driving? Well, so far, better than human. After all, in 250,000+ miles, they’ve only been in 2 minor fender-benders, one of which occurred while the human backup driver had the wheel, Google contends. Though some skeptics have their doubts.

Want to know more about Google’s autonomous car? Come back next Monday (6/25) for part 2 of our driverless car series.

Sound off

So now that you’ve got the lowdown, what do you think of driverless car technology? Does the prospect frighten or awe? Let us know!

References

Back to basics: how Google’s driverless cars get around

no comments yet

Leave a Reply

Leave your opinion here. Please be nice. Your email address will be kept private.