Why Google’s Fully Autonomous Car Won’t See Production Soon

Continuously evolving driving aids like brake assistance, adaptive cruise control and blind spot monitoring could make you believe fully autonomous cars are just around the corner. And Google took the lead earlier this year in the driving automatization with its cute steering-wheel-less blob on wheels that can carry two people around the streets. Well, as it turns out, the company’s concept has quite some flaws.
Google autonomous car 1 photo
Photo: Google
According to a report from Slate, the technology used by Google to make its demo fully autonomous car hasn’t got any chances to succeed in the real world. That would require an immense technological breakthrough that will endow computers human-like intelligence.

Thought the Google car navigates obstacles because it’s smart?

For those who skipped this chapter, they are wrong. Google’s autonomous car doesn’t navigate its way like a human being. That would require general intelligence, common sense and situational awareness, things current computers lack.

The little car actually relies a lot than you think on maps to go places. You might think that Google is quite an authority when it comes to mapping, but those maps are just a general guide for the computer.

For a more accurate mapping, Google has actually sent cars and people on the streets around the Mountain View headquarters in California (the only streets that the car has ever driven on) to map everything from poles to traffic lights and road signs.

This way, the car has a general idea what it will encounter and it’s sensors/cameras will check the surroundings for the objects already mapped. For example, it will know there’s a street light at the next intersection and it will scan the area where the device is to check if it’s green or red. And even that might not work properly if the sun “blinds” the cameras.

Without those complex maps, the Google autonomous car wouldn’t even budge.

The big picture

The G-mobile may have autonomously driven over 700,000 miles (1,1 million km) but all that distance was achieved going over and over again on the same streets around Google’s headquarters. For the big run, every foot of USA’s 4 million miles needs to be in-depth-scanned, including driveways, side streets and off road trails.

Let’s say this could be done in 10 years or so, but then comes an even bigger problem - map updates. The autonomous car lacks our intelligence and awareness, it’s only working with “true” or “false” algorithms and thus it won’t know what to do in unforeseen scenarios like road constructions or newly installed traffic lights/signs.

Even more of a problem is the fact that its sensors cannot make the difference between a rock laying in the middle of the road and a crumpled newspaper. So imagine the fun you’re going to have stopping or avoiding harmless obstacles.

And to top it off, the Guglewagen doesn’t know how to find parking spots. It could slowly go along a row of cars parked near the curb and scan for an empty spot like current cars do, but that doesn’t work in huge supermarket or multi-level parking lots.

Trying to fit cars artificial intelligence in order for them to be “aware” of their surroundings seemed the way to go, but until computers will be able to learn and act like human brains do, fully autonomous cars are still a long way dream.

Which is nice to hear if you’re an oldschool driver who enjoys his time behind the wheel. There’s still a “threat” though; Toyota has a plan to make cars communicate with each other and the infrastructure to make their way around with less or no human input. Still, that’s another idea that will take a lot of time to be implemented.
If you liked the article, please follow us:  Google News icon Google News Youtube Instagram

Would you like AUTOEVOLUTION to send you notifications?

You will only receive our top stories