All Sections

Google driverless cars not that smart, says Google driverless cars chief

Autonomous cars are slowly but inexorably headed to our streets, and while they clearly have many benefits, not everybody is convinced they’re as useful, smart or safe as some would have us believe – and that includes the people that actually make the things.

Google Car team director Chris Urmson has highlighted several key flaws in driverless car technology that, at best, might limit their use and, at worst, make them potentially dangerous to pedestrians.

Driverless cars: A distaster witing to happen?

While Google’s driverless car can detect pedestrians, it is currently unable to detect people standing at the side of the road, so police or lollipop ladies/men instructing cars to stop from the pavement will be completely ignored.

Urmson claims that, despite having been tested extensively across 700,000 miles during development, driverless cars are unable to navigate 99 percent of US mainland.

The problem, Urmson says, lies with the fact that driverless cars rely heavily on pre-set routes programmed into maps. Once on a route, the cars are unable to deviate of their own accord. If faced with a problem not identified in advance of its departure – such as road works or diversions – it would become confused.

There are other potential pitfalls, even if the route is clear. Sensors can mistake inclement weather, such as heavy rain or snow, for obstacles. Even bright sunshine can cause problems, with sensors potentially mistaking bright light for traffic signals.

Self-driving cars will completely ignore potholes, too, charging brought them at full speed, potentially causing damage.

Driverless cars aren’t quite ready for primetime, then, but Urmson remains confident these flaws can be ironed out in as little as five years.

Would you trust a robot car? Let us know in the comments below.

Comments