Join us on Zoom: https://ucsc.zoom.us/j/97555513871?pwd=NjhCaG9YYlFzOC9JbnFQZFhBcjlKQT09 / Passcode: 211111
Description: Scenario-based testing of autonomous vehicles (AVs) in a virtual environment reveals possible critical scenarios in the real world. In this work, we focus on a specific set of real-world scenarios regarding jaywalking and lay out our plan to create such scenarios using adversarial strategies, which can ensure the variety and representativeness of such scenarios. Our research intends to build a robust jaywalker model and an adversarial environment generator. Together these two tools are put into simulation against a target AV planner to fool their jaywalker movement prediction, thus creating critical scenarios.