Research papers · Posted 2024.10.11 · Preprint

AdvDiffuser: Generating Adversarial Safety-Critical Driving Scenarios via Guided Diffusion

Yuting Xie, Xianda Guo, Cong Wang, Kunhua Liu, Long Chen · arXiv · English

Overview

AdvDiffuser learns plausible background traffic through diffusion and uses a lightweight guide for adversarial behavior. Experiments assess realism, diversity and transfer to different driving systems with limited warm-up data.

Version: v1

AdvDiffuser: Generating Adversarial Safety-Critical Driving Scenarios via Guided Diffusion

Overview

Participation enquiry

Your enquiry has been sent.

AdvDiffuser: Generating Adversarial Safety-Critical Driving Scenarios via Guided Diffusion

Overview

Related resources

LLM-Guided Transformation of Non-Critical Driving Scenes into Safety-Critical Scenarios Using Augmented Reality

OmniSCS: Omni Safety-Critical Scenario Synthesis for Autonomous Driving via a Fully Editable Driving World

Generating Realistic Safety-Critical Scenarios for Vehicle-Pedestrian Interactions