Generating Potential Accident Scenarios for Autonomous Vehicles Using AI - NVIDIA DRIVE Labs Ep. 27

Опубликовано 23 февраля 2023, 17:00

Testing autonomous vehicles (AVs) in potential near-accident scenarios is critical for evaluating safety, but is difficult and unsafe to do in the real world. In this episode of DRIVE Labs, we discuss a new method from NVIDIA researchers called STRIVE (Stress-Test Drive), which automatically generates potential accident scenarios in simulation for AVs. The key to STRIVE is a generative AI model of traffic trajectories that encourages scenarios to be realistic even when they contain a potential accident. STRIVE produces diverse and plausible potential accident scenarios based on real-world data, allowing us to safely evaluate the planning part of the AV stack at scale in simulation.

00:00:00 - Adversarial Scenario Generation
00:00:05 - Why do we need to generate scenarios automatically?
00:00:20 - Overview of the STRIVE (Stress-Test Drive) method
00:00:40 - Creating potential accidents and solutions with STRIVE
00:01:05 - Modifying a real-world scenario in DRIVE Sim
00:01:51 - Ensuring realistic scenarios using generative AI traffic model
00:02:11 - Find out more

Read more: nvda.ws/3Isup58
GitHub: github.com/nv-tlabs/STRIVE
Watch the full series here: nvda.ws/3lHQP7H
Learn more about DRIVE Labs: nvda.ws/36r5c6t
Follow us on social:
Twitter: nvda.ws/3LRdkSs
LinkedIn: nvda.ws/3wI4kue