Audi Q2 deep learning - a step towards autonomous parking

Source: https://www.audi-mediacenter.com/en/photos/album/audi-q2-deep-learning-concept-728

Parking can be such as a frustrating chore and make you pull your hair out. Automakers have addressed it with a suite of automatic parking technology. But, what if the car can learn the art of parking overtime that it becomes so smart you can leave the parking chore wholly to the car? Enter Audi's Q2 deep learning concept.  

How does the concept work? In a nutshell, a pair of mono cameras are fitted front and rear. 10 ultrasonic sensors are placed at points all around the concept. An on-board computer will receive the information and converts the information into signals to control the steering and electric motor.

Source: https://www.audi-mediacenter.com/en/photos/album/audi-q2-deep-learning-concept-728

The concept will then scan for an available parking space by itself. For the purpose of this demo, the parking space is represented by a metal frame as shown in the image above.

After identifying where it is relative to the parking space, it will determine how it's going to navigate itself into the parking space and steer, drive forward and drive backward until it's perfectly parked inside the space. All this is done without human input.

How does deep learning algorithm enables full autonomous parking ability? Through good ol' trial and error. The algorithm will experiment with different number of manoeuvre until it identifies which manoeuvre will increase chances of success. Once it does, you may enjoy a hassle-free parking. The car will do it for you while you delightfully continue sipping your latte.

With the future seemingly heading towards an era of automation, this deep learning concept will, in my opinion, become a viable technological solution to create a more pleasant commuting experience.