Earlier this year, Audi sent an A7 Piloted Driving Concept on a 550-mile journey from San Francisco to Las Vegas and brought a group of journalists along for the ride. However not everything went to plan, as explained in an Automotive News piece of the problematic test.
Audi was expecting its autonomous A7 to complete the test without any issues, however ‘Jack’, as engineers called the concept, got a piece of tumbleweed lodged in its grille, blocking up one of the sensors. The car was able to stay on course, however the test highlights a common problem automakers face when developing self-driving vehicle tech.
Jen-Hsun Huang, CEO of software company Nvidia Corp., explained companies “can’t program a car to do every behavior.” That’s why they are developing systems that understand ‘deep learning’, which will train computers to think more proactively, like a human brain. These systems would take data from a camera or sensor and relate them to a series of potential driving situations, choosing the best course of action to take. These systems would learn continuously the more they were used and could also be updated to be better and better as time goes on.