Enhancing AV Simulation with Generative AI - DRIVE Labs Ep. 36

Опубликовано 21 августа 2024, 20:34

Simulation is critical to the development of safe autonomous driving systems. However, creating realistic scenarios that accurately depict rare and potentially dangerous situations can be extremely labor intensive. In this episode of DRIVE Labs, we discuss three key advancements from NVIDIA that use generative AI such as text-to-simulation to create realistic environments, generate natural driving behaviors, and edit the resulting scenarios to enable rigorous AV evaluation and training.

00:00:00 - Simulation is critical to safety of autonomous driving systems
00:00:42 - Building realistic road layouts with MapLLM
00:01:18 - Generating natural driving behaviors with LCTGen
00:02:20 - Editing the result with Scenario Editor
00:03:36 - Future integration into the NVIDIA NIM roadmap

Resources:
Paper: Language Conditioned Traffic Generation: ariostgx.github.io/lctgen
Paper: RealGen: RAG for Controllable Traffic Scenarios: arxiv.org/abs/2312.13303
Video: AV Development With Foundation Models, Dr. Marco Pavone, GTC 2024: nvidia.com/en-us/on-demand/ses...
Video: NVIDIA AI Tools for AV Developers, GTC 2024: youtube.com/watch?v=LLSuUBObtt...

Watch the full DRIVE Lab series: nvda.ws/3LsSgnH
Learn more about DRIVE Labs: nvda.ws/36r5c6t

Follow us on social:
Twitter: nvda.ws/3LRdkSs
LinkedIn: nvda.ws/3wI4kue

#NVIDIADRIVE