Advanced Lane Detection Model for the Virtual Development of Highly Automated Functions

Authors

T. Kanuric BSc, H. Li  MSc, Assoc. Prof. Dipl.-Ing. Dr. techn. A. Eichberger, Institute of Automotive Engineering, Graz University of Technology; Dr. techn. Dipl.-Ing. D. Nalic, MQS Automotive GmbH & Co KG, Heilbronn; P. Pannagger MSc, Mercedes-Benz Group AG:

Year

2023

Print Info

Production/Publication ÖVK

Summary

Virtual development and prototyping have already become an integral part of automated driving systems (ADS). Essential components of this development in the context of automated systems are the development of real-time capable sensor models. This paper presents a sensor model and its implementation in a real-time validation framework. For this purpose, a neural network-based phenomenological lane marking detection model is developed and integrated into a real-time virtual vehicle testbed system. The proposed lane marking model aims to offer a realistic and real-time capable sensor model that can be used to evaluate and develop lateral path guidance systems. In addition, the complete simulation framework provides a complex virtual environment and an accurate vehicle dynamics model. The results demonstrate the performance of the calibrated and developed lane marking model. Meanwhile, the simulation framework applied also provides the most flexibility for further expansion of Vehicle-in-the-Loop in the future.

ISBN

978-3-9504969-2-5