MIT and Motional create CW-Net to explain autonomous vehicles
MIT and Motional have teamed up to create CW-Net, an AI tool that helps self-driving cars explain their actions in plain language: think "approaching stopped vehicle" or "close to cyclist."
The goal? To make riding in autonomous vehicles feel safer and more understandable for everyone.
CW-Net provides real time insights
CW-Net gives real-time insights into what the car's AI is thinking, so safety drivers and passengers can spot unsafe moves before they happen.
During tests, it helped engineers identify and address problems in the vehicle's AI system, including emergency braking after getting too close to a cyclist.
With training on 130 million driving scenes, CW-Net works with existing systems without slowing them down, and researchers are planning to expand its abilities even further.