U.K. startup Wayve said it has created an autonomous vehicle that can learn to drive the way a human does, through trial and error, using cameras and satellite navigation.
The company said this "is only possible with end-to-end machine learning."
With a backup driver in the car to take over should the vehicle be in error, this approach enables the automobile to learn by experience, mistakes, feedback, and imitation.
Said Wayve co-founder Alex Kendall, "Every time a safety driver intervenes or takes over control of the vehicle, we learn from that experience and that feedback. And with each piece of data, we're able to train our system to get better and better and better."
The company says its approach renders unnecessary more sophisticated training technologies and techniques, like high-definition maps, sensor/computer suites, and hand-coded rules.
From The Washington Post
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA