Safety-Critical Traffic Simulation with Guided Latent Diffusion Model

2025-05-01

AI Summary

Existing safety-critical traffic simulation methods suffer from insufficient physical plausibility and low generation efficiency, hindering robustness evaluation of autonomous driving systems. To address this, we propose the first graph-structured variational autoencoder (VAE)-guided latent diffusion framework tailored for generating hazardous rare scenarios. Our method comprises three key components: (1) a graph VAE that explicitly models multi-agent interactions and topological relationships; (2) differentiable physics-based constraints—encompassing kinematics, collision avoidance, and interaction plausibility—integrated into the latent diffusion process to ensure physical fidelity; and (3) a feasibility-driven posterior filtering mechanism to enhance sample quality. Experiments on nuScenes demonstrate significant improvements: +28.6% adversarial triggering rate and 3.2× speedup in generation efficiency, while achieving state-of-the-art physical plausibility and behavioral realism.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Diffusion Models for Vision
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Safety-critical traffic simulation plays a crucial role in evaluating autonomous driving systems under rare and challenging scenarios. However, existing approaches often generate unrealistic scenarios due to insufficient consideration of physical plausibility and suffer from low generation efficiency. To address these limitations, we propose a guided latent diffusion model (LDM) capable of generating physically realistic and adversarial safety-critical traffic scenarios. Specifically, our model employs a graph-based variational autoencoder (VAE) to learn a compact latent space that captures complex multi-agent interactions while improving computational efficiency. Within this latent space, the diffusion model performs the denoising process to produce realistic trajectories. To enable controllable and adversarial scenario generation, we introduce novel guidance objectives that drive the diffusion process toward producing adversarial and behaviorally realistic driving behaviors. Furthermore, we develop a sample selection module based on physical feasibility checks to further enhance the physical plausibility of the generated scenarios. Extensive experiments on the nuScenes dataset demonstrate that our method achieves superior adversarial effectiveness and generation efficiency compared to existing baselines while maintaining a high level of realism. Our work provides an effective tool for realistic safety-critical scenario simulation, paving the way for more robust evaluation of autonomous driving systems.

Problem

• Generating realistic rare traffic scenarios for autonomous driving tests
• Improving physical plausibility and efficiency in scenario generation
• Enabling controllable adversarial behaviors in simulated traffic scenarios

Innovation

• Guided latent diffusion model for realistic scenarios
• Graph-based VAE captures multi-agent interactions
• Physical feasibility checks enhance plausibility