GTC 2020: Virtual Validation of Automated Vehicle Safety: Challenges, Lessons Learned, and Opportunities
After clicking “Watch Now” you will be prompted to login or join.
Click “Watch Now” to login or join the NVIDIA Developer Program.
Virtual Validation of Automated Vehicle Safety: Challenges, Lessons Learned, and Opportunities
Jim Cherian, CETRAN (Center for Testing and Research on Autonomous Vehicles - NTU), Nanyang Technological University, Singapore
We'll explain the need for virtual validation (VV) of automated vehicles (AV), identify the general strengths and weaknesses of VV, briefly review the present landscape of tools and solutions available to perform AV simulations, and list the major building blocks of a good virtual testing process (VTP). We'll also discuss the potential of NVIDIA DRIVE Constellation toward implementing a meaningful VTP for developers, regulators, and service providers. We'll outline some key challenges and opportunities to motivate further research. We'll also introduce two research projects at CETRAN (Nanyang Technological University, Singapore): first, modeling the sensing and perception (S&P) errors using a Perception Error Model (PEM) and studying their impact on decision-making (AV behavior), and second, developing methods to evaluate the holistic fidelity of simulation toolchain and judge its effectiveness in representing reality. We'll present some preliminary findings and a research roadmap.