'Hocus Pocus 2' Review Wi-Fi 6 Router With Built-In VPN Sleep Trackers Capital One Claim Deadline Watch Tesla AI Day Student Loan Forgiveness Best Meal Delivery Services Vitamins for Flu Season

Audi's new concept taught itself how to park

Using algorithms and deep learning, this scale-model Q2 can probably park better than you.

Audi Q2 deep learning concept , model car on a scale of 1:8

It's always unnerving to see computers "learn" how to do things. Thankfully, the Audi that taught itself how to park is a small-scale model, so it won't be starting a machine uprising any time soon. Hopefully.

This is the Audi Q2 Deep Learning Concept. It's a 1:8 model built with just one purpose in mind -- learning how to park. Placed on a metal frame, it can find the designated parking spot and slide itself right in without any human intervention.

The model might be 1:8 scale, but the ultrasonic sensors are very clearly still full-size.


The system uses 10 ultrasonic sensors and two cameras. It finds the space, calculates how it needs to get into said space, then it starts maneuvering. It learns through a type of trial and error, continually refining its movements based on successful parking attempts. Audi's engineers will eventually scale this up to a proper production model.

This is all part of Audi's future-tech subsidiary, Audi Electronics Venture, based in Germany. Audi plans to flesh out its deep-learning software for environmental perception with the help of Mobileye, a firm leading the way in image recognition. It's also relying on the help of Nvidia, which helped develop the hardware controller for this system.

Audi said it will implement this hardware controller in the next generation of its A8 full-size luxury sedan. The car likely won't be capable of teaching itself how to park from scratch, but it sounds like it will be able to utilize deep learning to improve its processes.

This type of artificial intelligence will be necessary as the industry introduces autonomous cars. Self-driving vehicles need to be able to read the road ahead and evaluate any situation that may arise. You can't hard-code a lifetime of human experience into a car's computer, so deep learning and AI is basically a requirement for taking the next step in automotive technology.