Stanford's autonomous Audi hits the track to improve its AI safety

While Audi already has its own, err, self-driving Audi, researchers at Stanford University have been working with an autonomous TTS for several years now, developing their own artificial intelligence. Now the school has shared that they're even hitting the race track, letting their car navigate the course at speeds up to 120mph. The purpose is not to have the Audi TTS, dubbed "Shelley," compete in a circuit with real humans, but rather to use the data from high speed driving to improve safety.

Chris Gerdes, a Stanford professor involved in the project, explains that "race car drivers can use all of a car's functionality to drive fast... We want to access that same functionality to make driving safer." While one benefit of completing high-speed laps is collecting data on their AI can best handle corners, they also have Shelley spend most of the time driving between 50 and 75mph, the speeds with the highest rates of accidents occurring.

This allows the students to focus on improving algorithms that would be vital during situations such as emergency lane-changes or driving on the freeway. Braking, throttle, and maneuvering are all factors that play a part in avoiding collisions, so Stanford's software needs as much data as possible on effectively using all three.

Of course, all this doesn't mean Shelley hasn't gotten really good at setting lap times. The researchers say the car is now "almost as fast around the track as an experienced racer." One student recently recorded a human driver on the course, with the goal of incorporating their behavior into the AI's algorithms, hopefully one day allowing autonomous vehicles to be as skilled and safe as professional drivers.

SOURCE Stanford University