Audi opens centre for self-learning autonomous vehicles

Posted on 26 Jun 2017 by Jonny Williamson
The Audi JKU Deep Learning Centre supports the manufacturer to produce self-learning cars which behave more safely in complex traffic situations - image courtesy of Audi.

Audi has announced that Austria's Johannes-Kepler-University Linz (JKU) will help support the development of AI-based technologies for improving the self-learning functions of its future autonomous vehicles.

autonomous vehicles - The Audi JKU Deep Learning Centre supports the manufacturer to produce self-learning cars which behave more safely in complex traffic situations - image courtesy of Audi.
The Audi JKU Deep Learning Centre supports the manufacturer to produce self-learning cars which behave more safely in complex traffic situations – image courtesy of Audi.

The newly launched Audi JKU Deep Learning Centre aims to further the implementation of automotive artifical intelligence (AI) technologies and supports Audi’s aim of creating self-learning autonomous vehicles capable of navigating complex traffic situations more safely.

One project of the programme focuses on how future autonomous vehicles can better recognise intricate situations at an early stage and what must be done to optimise reaction times.

To achieve this technical accuracy, the correct implementation of vehicles sensors is one essential field of research within the programme. Precise vehicle localisation is crucial, according to JKU, since autonomous vehicles can only be exactly aware of and interpret the complex environment they are moving in, when they are 100% certain.

A second project deals with speech recognition technologies, which enables drivers to better communicate with their cars.

This approach is based on Long Short-Term-Memory (LSTM), a recurrent neural network architecture, proposed back in 1997 by Prof Sepp Hochreiter, an AI expert at Linz Uni.

The LSTM is already being employed for most versions of voice recognition software in mobile devices, and is likely to determine the AI future even more firmly when it comes to developing technologies for new autonomous driving industries.

Hochreiter commented: “There is still a huge development potential lying in the implementation of AI methods for self-learning autonomous vehicles. It is not only that voice recognition will enable a wireless communication between the driver and the car. Furthermore, AI technologies will facilitate the cars themselves to predict the drivers’ intentions and wishes and the cars will learn how to react appropriately to those.”

Hochreiter added, that thanks to LSTM, future drivers would feel much more comfortable in their cars than in their living rooms.