ATZ worldwide

, Volume 121, Issue 12, pp 42–45 | Cite as

Artificial Intelligence in the Development of Autonomous Driving Functions

  • Natalya AhnEmail author
  • Raphael Pfeffer

The automotive industry can significantly benefit from artificial intelligence applications - especially in the development of advanced driver assistance systems and autonomous driving functions. Integrating the relevant algorithms, however, may entail a high effort, says IPG Automotive.

Ensuring data quality

At a fundamental level, a learning process of deep neural networks requires large volumes of training data such as images. In supervised learning, these need to contain meta-information such as the position and classification of all relevant objects in addition to the actual images. Usually, these training data are labeled manually or in semi-automated mode, which entails a very high effort. Moreover, a large amount of the training data is required with high variance to ensure that as many conceivable environmental situations as possible are covered.

The quality of the training data contributes significantly to the quality of the trained neural network. Neural networks will only...

Copyright information

© Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2019

Authors and Affiliations

  1. 1.IPGKarlsruheGermany

Personalised recommendations