Cambridge Startup Teaches Driverless Cars To Behave Around Cyclists, Can Deploy On Any Road Now

“Cambridge is a great place to test [autonomous vehicles] because the roads are complex and difficult,” says the cofounder of driverless car startup Wayve.

Alex Kendall adds:

Each hour the streets are flooded with students cycling between lectures. This is great for challenging us to build a complex AI [artificial intelligence] system. Other approaches, focusing on easy cities like Phoenix, Arizona – with sunny, wide grid roads – will never scale to these situations. In Cambridge, we’re forced to develop something which is truly intelligent.”

Wayve is attempting to disrupt the multibillion-dollar autonomous vehicles (AV) industry by not relying on arrays of LIDAR sensors, 3-D HD mapping, or hard-coded how-to-drive rules. Instead, it uses machine-learning delivered via basic cameras, a smartphone for directions, and an onboard computer no more capable than a standard laptop.

Engineers at the startup teach their software basic driving lessons in a simulated environment (one of the firm’s current job ads is for a gaming designer), and then, in the real world, human drivers take over when Wayve cars make Errors.

With this literally hands-on approach a Wayve driverless car can be taught how to drive around busy and narrow roads in a matter of hours, claims the firm.

Simpler tasks would be easier for the Wayve algorithms to conquer.

Quelle: Read more

Veröffentlichung:
16. April 2019