Data-Driven Generation and Analysis of Hazardous Traffic Scenarios

ABOUT THIS PROJECT

At a glance

Urban intersections are the most challenging operational design domains for the automated vehicles. The challenge stems from complexity of human and robotic interactions as well as temporal and spatial uncertainties that may lead to collisions. Over the last two years, our team has produced a workflow that generates specification of challenging traffic scenarios from data using a new programming language called Scenic, connected to back-end simulators. In this new project, we plan to significantly extend our work by incorporating agent dynamics into Scenic, allowing for specifying more complex agent behaviors, and developing sophisticated causality analysis methods to understand the reasons for hazardous/unsafe scenarios based on our simulation-based analysis.

This project builds upon results from "Data-Driven Synthesis of Hazardous Scenarios at Traffic Intersections". 

PRINCIPAL INVESTIGATORSRESEARCHERSTHEMES

Sanjit A. Seshia

Pravin Varaiya

 

simulation, verification, hazard analysis, causality, traffic intersections, intelligent infrastructure