As automakers continue to work on autonomous features in their cars, there have been many self-driving features being added to the cars. Audi too is looking to train and learn from the autonomous features as it looks to bring the technology to its cars.
The company says the next step will be machine learning-powered driver assist features which work during a pressurising situation like parking and driving in traffic. This feature will soon be available in the Audi A8 saloons in 2017. But for now, Audi is testing these features in an ⅛ the size of the Q2 SUV model toy car.
In the video above, you can see the toy car in action using deep reinforcement learning - a type of machine learning which is essentially trial and error with scores for the computer when it gets anything right in the training.
Audi's engineers have enabled the toy car with the autonomous feature to search and park in a 3x3 meter arena. The parking by the model toy car is simulated with several million repetition and the algorithm rewards for correct parking.
Also, the Audi engineers check if the model toy car can park from different position as well for this they also place the identification marker randomly.
Audi in a blog post said that an algorithm identifies the successful actions, thus continuously refining the parking strategy. So, at the end, the system is able to solve even difficult problems autonomously.
Audi is not the only player in the auto industry to use machine learning to develop autonomous driving, but is one of the biggest and the technology will affect many drivers.
In the video, Audi trains the autonomous software to park the car and learn on its own. The model toy car is connected with different ultrasonic sensors along with cameras.