A British startup says it doesn’t need the expensive sensing tech found in most self-driving cars and has opted instead to use cameras and artificial intelligence software that learns how drivers react to real-world situations
Medium reported Monday that Wayve fits its test cars with a camera, a GPS tracker, and a computer to analyze collected data. Unlike other self-driving cars, the Wayve vehicle lacks costly lidar and radar systems.
Instead, the Wayve car’s computers are expected to use cameras to learn from a driver’s real-world reactions. Wayve’s approach is to plug what the company calls “reinforcement” data gleaned every time a human driver makes a correction. Each mistake is a lesson learned for the car, and eventually Wayze believes the data will help it avoid the error.
With this strategy, the company recognized it’s not on the same development curve as every other startup and admitted it will take longer to reach its first autonomous vehicle deployed on the streets.
Wayve’s less is more approach doesn’t appear like a winning strategy on paper, said Rick Tewell, Velodyne LiDAR chief operating officer. Tewell told Medium that artificial intelligence always works better with more data available, not less. Without radar or lidar, the self-driving car is effectively handicapped for information other autonomous cars are fed on the regular. Wayve argued its setup and costs are 10 percent of what major self-driving car rivals might spend.
For automakers with billions to funnel into the budding segment, cost may not be as pressing as urgency.