Title:

Injecting Traffic Rules in AI-Powered Autonomous Vehicles

Abstract:

This talk explores the injection of traffic rules into AI-powered autonomous vehicles to improve their safety and compliance to the traffic code. Broadly, the talk is structured in two parts: The first focuses on a methodology for a parsimonious yet highly expressive encoding of traffic rules using the language of Signal Temporal Logic (STL). The second part of the talk discusses approaches to leverage this encoding for traffic rule injection in the AV stack using tools from out-of-distribution (OOD) detection and Bayesian filtering.

Bio:

Sushant Veer is a Senior Research Scientist with the Autonomous Vehicle Research Group at NVIDIA Research. His current research interests lie in developing algorithmic guardrails for AI-driven autonomous vehicles to ensure their safe deployment and to facilitate safety verification and validation. In the past, he has worked on the development of provably safe motion planners for drones and dynamic-legged robots and robust control of uncertain hybrid dynamical systems. Before joining NVIDIA, he was a Postdoctoral Research Associate in the Mechanical and Aerospace Engineering Department at Princeton University. He received his Ph.D. in Mechanical Engineering from the University of Delaware in 2018 and a B.Tech. in Mechanical Engineering from the Indian Institute of Technology Madras in 2013.