Update: Cancelled due to technical issues

ITSS Distinguished Lecture by Dr. Reza Rezaei


Virtual Field Development and Testing Methodology for AD/ADAS System

Brief Abstract

The development of autonomous driving systems requires comprehensive testing and validation to ensure that the regulatory requirements are fulfilled, and the system performs reliably and safely under all operating conditions.  Innovative virtual field validation methods are becoming increasingly important not only due to the cost reduction in the development but also for covering more operating conditions. It also allows for testing of rare and dangerous scenarios without risking physical harm to the test vehicle, and traffic participants.

The current results of virtual field testing and validation methodology of AD/ADAS systems, with a specific focus on visual perception will be discussed. First, an overview of the AD/ADAS development process and the use of virtual testing are presented. Then the challenges associated with camera-based perception systems and computer vision algorithms under various environmental and lighting conditions, including camera soling are discussed.

The model-based methodology developed to create adverse and challenging scenarios for computer vision systems is demonstrated with multiple simulation results based on real-field measurement data. Finally, an outlook on future developments of the AI-based methodology for creating critical scenarios and edge cases using deep reinforcement learning is provided.

Brief Bio

Priv.-Doz. Dr.-Ing. habil. Reza Rezaei is Manager for Modeling and Simulation of Intelligent Perception Functions for Autonomous Driving at IAV GmbH in Germany. In parallel, he is a guest lecturer at the Leibniz University of Hanover and adj. Prof. at the University of Alberta in Canada with focus on Artificial Intelligence and mechatronic systems. He has a track record of fundamental research on these topics documented by numerous publications by IEEE and SAE.