Friday, October 28, 2022 | 12pm to 1pm
About this Event
Speakers: Prof. Daniela Rus & Dr. Alexander Amini
Host: Prof. Jinhua Zhao, MIT
Abstract: Integrating autonomy deep into society could revolutionize urban mobility towards a more efficient, safe, and sustainable future. This vision of mobility requires new technological solutions to safeguard against the unexpected events and harsh uncertainties of our rapidly changing world. In this talk, we present new data-driven approaches towards achieving resilient, dynamic, and adaptable mobility systems. We share new methods for computationally designing high-fidelity, synthetic environments directly from real world data to simulate a continuum of scenarios that are too harsh or difficult to collect in reality. This simulation platform, VISTA, provides a open-source, powerful, and flexible playground for sustainably collecting rich data and training intelligent models across a variety of perturbations, multi-agent interactions, and environmental conditions. Furthermore, to proactively protect systems into deployment, we present modular methods to model and anticipate key uncertainties, providing a step towards resilient and adaptive decision making.
Background reading: VISTA 2.0: An Open, Data-driven Simulator for Multimodal Sensing and Policy Learning for Autonomous Vehicles (pdf)