AI controls the amount of data for automated driving

Joint project KIsSME with KIT participation starts - learning algorithms enable efficient recording of scenarios when testing highly automated vehicles

Automated vehicles use various sensors to record information about themselves and their surroundings. (Photo: Markus Breig, KIT)

Automated vehicles use various sensors to record information about themselves and their surroundings. (Photo: Markus Breig, KIT)

When testing highly automated vehicles, large amounts of data are generated. The new KIsSME project aims to reduce this amount of data in order to save storage space, electricity and evaluation effort, but at the same time to condense the information in order to make the vehicles safer. Algorithms based on artificial intelligence (AI) select the data while driving and sort it into scenario catalogs. The Karlsruhe Institute of Technology (KIT) provides data from driving tests and simulations in the joint project funded by the Federal Ministry of Economics. 

Highly automated driving promises many advantages - more comfort for the driver, fewer accidents, smoother and therefore more environmentally friendly road traffic. The vehicles of the future will have many different sensors with which they will receive information about their own status and from their surroundings. Using this information, they need to make reliable driving decisions in the shortest possible time. When testing, each vehicle variant has to cover millions of kilometers and master many different scenarios that combine infrastructure, weather conditions and other road users and their behavior. "This generates huge amounts of data - four to eight terabytes per vehicle per day," reports Dr. Michael Frey, Deputy Head of Institute at the Institute for Vehicle System Technology (FAST), Vehicle Technology Branch of KIT. "These amounts of data can hardly be handled anymore." 

Intelligent data selection algorithms

Creating catalogs of driving scenarios and sorting out newly occurring scenarios during vehicle testing, i.e. recording only those data that actually add value while driving - that is the approach of the joint project KIsSME (which stands for "Artificial intelligence for selective near real-time recording of scenario and maneuver data when testing highly automated vehicles ”). To this end, scientists are developing AI-based algorithms that select the data that is generated while the vehicle is in motion. "KIsSME aims to expand the scenario catalog and at the same time reduce the amount of data," explains Frey, who among other things heads the "Automation" research group at FAST.

Tests are running on the test field autonomous driving Baden-Württemberg (TAF BW)

The KIT researchers provide data from real driving tests and simulations for KIsSME. For this purpose, test drives are carried out in public urban traffic and on the test field Autonomous Driving Baden-Württemberg (TAF BW) in Karlsruhe as well as closed vehicle-in-the-loop simulations on a complete vehicle test bench at KIT. FAST researchers also check the AI models and AI selectors developed in the project by applying the algorithms developed by the network partners to the data from tests and simulations. KIsSME refers to automated driving from levels four to five (fully automated to autonomous).

The coordination of the joint project KIsSME lies with AVL Deutschland GmbH. In addition to KIT, the Fraunhofer Institute for Short-Term Dynamics, the Ernst Mach Institute, the FZI Research Center for Information Technology, LiangDao GmbH, Mindmotiv GmbH, RA Consulting GmbH and Robert Bosch GmbH are involved as partners. ASAM eV, Association for Standardization of Automation and Measuring Systems, and the Cluster Electromobility South-West, coordinated by e-mobil BW GmbH, State Agency for New Mobility Solutions and Automotive Baden-Württemberg, act as associated partners.

More about Karlsruhe Institute of Technology

author avatar

KIT is one of the biggest research and education institutions worldwide and has the potential of reaching a top position in selected research areas on an international level. The objective is to turn KIT into an institution of top research, excellent scientific education, and a prominent location ... learn more

Stay Informed, Be Inspired

Wevolver’s free newsletter delivers the highlights of our award winning articles weekly to your inbox.