Last week, Audi attended the Conference and Workshop on Neural Information Processing Systems (NIPS) in Barcelona, one of the world’s most important specialist conferences for artificial intelligence. The Bavarian brand was at the event for the very first time to showcase its expertise, and it did so in the form of a model Audi Q2 that’s able to autonomously master complex parking situations.
On a 3x3m area, a 1:8 scale model car dubbed the ‘Audi Q2 deep learning concept’ autonomously searched for and found a suitable parking space – and parked itself! The intelligent parking process shown at NIPS demonstrates how self-learning systems are a key technology for piloted driving cars.
How does it work? Two mono cameras and ten ultrasonic sensors provide control signals for steering and the electric motor. The model Q2 determines its position relative to the parking space and then calculates how it can safely drive to its targeted destination, manoeuvring, steering and driving forwards or reversing as required. Learning by trial and error, the car initially selects its direction of travel at random, but an algorithm autonomously identifies the car’s success, meaning that the parking strategy is continually refined. They say practice makes perfect and it’s true in this case, as the system is eventually able to solve even difficult problems autonomously.
The next stage of the ‘Audi Q2 deep learning concept’ will see developers move the autonomous parking process to a real car – we can’t wait to see what happens!