Sharing the World with Autonomous Systems: What Goes Wrong and How to Fix It (NSF)

Award Number: CNS-2211141
Amount: 599,854
Start Date: 06/15/2022
End Date: 05/31/2025
Collaborators: Ufuk Topcu (UT Austin)

As autonomous systems start to operate in open, uncontrolled environments alongside humans, safety becomes a major concern. In applications in which human-operated systems and autonomous systems are in close interaction, the heterogeneity causes different agents to exhibit different behaviors under the same situation due to the differences in how they see the world and make decisions. For example, autonomous vehicles tend to be more conservative than average human drivers, leading to instances of confusion and frustration of human drivers when encountering an autonomous vehicle. As a result, understanding the effects of inconsistencies among interacting agents on the overall system is critical for the adoption and acceptance of autonomous systems.

This project aims to develop novel algorithms that will quantitatively and succinctly characterize the sources and effects of inconsistencies in perception and decision-making among the interacting agents. Based on this characterization, the project will develop efficient synthesis algorithms for autonomous systems to actively identify the inconsistencies and influence the behaviors of the other agents to improve the safety of the overall system. A key innovation lies in a rigorous approach integrating probabilistic formal methods, games on graphs, joint perception and planning, and convex optimization. While the resulting algorithms will be applicable to a wide range of cyber-physical systems, demonstrations will leverage an experimental platform in which autonomous vehicles interact with each other as well as human-controlled vehicles in mockup urban environments. The project will likewise promote outreach to industry, regulatory agencies, and the broader cyber-physical system (CPS) community through technical short courses. All educational material, demonstrations and software generated by the project will be shared publicly.

Tichakorn (Nok) Wongpiromsarn
Tichakorn (Nok) Wongpiromsarn
Assistant Professor

My research focuses on formal methods, motion planning, situational reasoning, hybrid systems, and distributed control systems.