Can AI make self-driving cars fall in line with roads?

Jul 09, 2018, 8:16 AM EDT
(Source: Marc van der Chijs/flickr)
(Source: Marc van der Chijs/flickr)

Automakers are racing to train their self-driving vehicles to fall in line with the laws of roads but the ride has been bumpy. Now a British startup, Wayve, has come up with an alternative approach to teach vehicles to drive autonomously with an AI algorithm instead of sophisticated hardware and detailed 3D maps.

Wayve’s method requires just one camera mounted at the front of vehicle, with the real-time feed channeled to a graphics processing unit (GPU) onboard the car, reports Futurism. With trial and error multiple times, the car learns to navigate a specific set of road conditions within just 20 minutes.

If this method gains traction, the development of autonomous vehicles would sharply turn to software-based framework instead of existing model heavily reliant on external sensors and custom-built hardware, notes NewsBytes.