RELAI: Scenario-based training for automated vehicle

How can we be sure automated vehicles will do what they should out on the road? The Fraunhofer IAO and the Institute of Human Factors and Technology Management IAT at the University of Stuttgart are looking to answer that question. A joint research team has set out to develop training and testing scenarios that will ensure AVs live up to our expectations.

In the interest of traffic safety, automated vehicles (AVs) have to behave as we all – drivers, passengers and everyone else on the road – would expect. This is why researchers engaged in the RELAI project are collecting real-world traffic data. This information will provide the baseline for creating testing scenarios to determine how AVs respond, for example, when approaching a crosswalk. The idea is for the vehicular control systems to learn from these test scenarios so they ‘know’ what occupants and other road users expect the vehicle to do in certain situations. This teaches the vehicle to respond as expected.

First, the Fraunhofer IAO and the University of Stuttgart IAT are going to send a specially equipped vehicle out on the Baden-Württemberg Autonomous Driving Test Field to collect data in actual traffic. These observations go to investigate situations where vehicles interact with more vulnerable traffic participants, particularly pedestrians. The researchers will then map these situations to the virtual models that are going to underpin scenario-based AV testing and training. Follow-up studies will determine if the behavior learned from the test scenarios really does meet road users’ expectations.