Audi Q2 Deep Learning Concept
It took Audi nine months to grab their Q2 small SUV
and transform it into a 1:8 scale model that demonstrates an intelligent
parking process. Named the 'Audi Q2 Deep Learning Concept' and showcased at the
Conference and Workshop on Neural Information Processing Systems (NIPS) in
Barcelona, Spain, where it will remain until December 10, the study features
two mono cameras, facing forward and backward, and 10 ultrasonic sensors
positioned on key parts of its bodywork.
The data is sent to an on-board computer, which
converts it into control signals for steering and electric motor, enabling the
Q2 concept to calculate its targeted destination and steer and drive it forward
or in reverse, depending on the situation, into the clear parking space.
Created by Audi Electronics Venture (AEV), an Audi
AG subsidiary in Gaimersheim, Germany, the scale model learns through trial and
error and paves the road for the brand's next step, which will transfer the
parking-space search process to an actual car.
Audi is also working with key partners
from Silicon Valley, Europe and Israel to create various software and
artificial intelligence components that will be eventually introduced on
production cars. One of these partnerships sees Mobileye and NVIDIA working
together with Audi in a deep learning-based software for environment
perceptions systems that will be launched next year in the next generation A8's
central driver assist controller.
