AVL, TÜV SÜD and NVIDIA working to standardize validation of autonomous vehicles using NVIDIA DRIVE Constellation simulation
AVL, TÜV SÜD and NVIDIA are working to develop a simulation test suite and standardized procedures for autonomous vehicle validation. TÜV SÜD and AVL have created an initial set of testing measures for manufacturers of automated driving systems, running them on the NVIDIA DRIVE Constellation simulation platform.
Though traffic accidents happen daily, it is rare for an individual driver to encounter an accident. Autonomous vehicles need to be prepared for all eventualities. They need to perform better than human drivers; they must gather robust experience in even the most infrequent adverse weather and light conditions, as well as countless traffic and road situations.
It would take years to physically drive an autonomous vehicle the millions of miles on public roads that would be necessary before certification. It is only through a combination of physical tests and simulations that manufacturers can validate their vehicles over such large distances under limitless conditions, thus ensuring system robustness.
Due to their competencies, the contributions of all three partners complement each other in a toolchain: the accuracy of NVIDIA’s cloud-based simulation platform enables the certification logic of TÜV SÜD’s together with the use of an integrated validation toolchain supported by AVL.
AVL has developed the test bench AVL DRIVINGCUBE for a reliable and efficient validation of autonomous driving systems integrated in a physical vehicle. The vehicle with physical or behavior sensor models connects autonomous driving functions under test with the simulated virtual environment and validate them under defined driving scenarios in this vehicle-in-the-loop testing environment.