Testing simulated scenarios in real environments for autonomous driving
The primary objective of the project is to create an asset for fortiss in the field of Autonomous Driving (AD) by leveraging expertise from different competence fields of fortiss, including Machine Learning (ML), Automated Software Testing (AST), and Software Dependability (SD).
A key focus of this project is to establish a comprehensive approach for testing critical traffic scenarios, encompassing both small-scale and full-size vehicles. The initial phase in preparing for testing with a full-size vehicle involves optimizing the AD software stack within our research vehicle fortuna. The testing of a diverse set of simulated scenarios in real-world environments is then demonstrated. This improved AD software stack is also planned to be deployed on an embedded platform (i.e., Nvidia Jetson TX2) in alignment with the evolving trends in the automotive industry. Furthermore, the project endeavors to execute efficient runtime monitoring of AD functions within a simulation environment, a crucial step towards the safe and successful adoption of these functions.
Finally, a comprehensive white paper for our AD roadmap will be published. This paper encapsulates our collective experiences, distills existing technical solutions, identifies new research questions, and depicts our medium-term development plans for 2024 and 2025.
The main contribution of the project is to build and demonstrate an up-to-date baseline for testing a diverse set of simulated scenarios in real-world settings both with small-scale prototypes and full-size vehicles. In addition, the up-to-date AD software stack/module is planned to be run and evaluated on embedded platforms which serve as a quasi-series computing hardware for many automotive OEMs.
Further, the BARK framework is intended to be extended with an efficient runtime monitoring of the AD functions for the purpose of detecting and identifying potential failures in decision-making.
01.12.2023 - 30.04.2024
