Semantic-Level Conflict Traffic Scenario Generation Via Spatiotemporal Polygon Anchors
Yunwei Li, Anran Wang, Siyu Wu, Shengjie Fu, Shuo Feng, Hong Wang, Jun Li
AI summary
Problem
Existing traffic scenario generators rely on low-level controls that fail to capture high-level, event-specific semantic demands, limiting their ability to produce realistic, safety-critical scenarios for rigorous autonomous driving testing.
Approach
The framework introduces Spatiotemporal Polygon Anchors (SPAs) to encode geometric and temporal conflict patterns, automatically extracting them from data and applying them as differentiable guidance within a diffusion model alongside a dynamic resampling strategy.
Key results
- Superior semantic controllability and event satisfaction over diffusion and optimization baselines
- Automated extraction of discriminative SPAs from target versus normal scenario datasets
- Effective exposure of latent functional weaknesses in tested autonomous driving systems
- Balanced exploration and refinement via dynamic resampling during diffusion denoising
Why it matters
Enables rigorous, semantically aligned safety validation for autonomous driving systems by bridging abstract test requirements with physically plausible simulation.
Abstract
Autonomous Driving Systems (ADS) require rigor- ous and complex testing under diverse conditions to fulfill var- ious demands and purposes of testing tasks, such as occlusion- triggered events, necessitating semantic-level control in scenario generation. Existing methods, reliant on low-level state controls, struggle to represent high-level semantic intents for task- oriented testing. We propose SPATSG, a novel framework for event-driven, semantically aligned Traffic Scenario Generation, leveraging Spatiotemporal Polygon Anchors (SPA) to bridge high-level test requirements and low-level diffusion guidance. SPAs encapsulate critical geometric and temporal patterns of traffic agents, derived from a set of targeted scenarios. During diffusion denoising, SPATSG integrates SPAs via an auxiliary loss to steer sampling toward desired semantics. A dynamic resampling strategy further intensifies guidance and prioritizes promising trajectory candidates progressively to balance exploration and refinement. We evaluate SPATSG on SinD, a Chinese intersection benchmark featuring complex interactions and diverse conflicts. Experiments on occlusion- triggered scenario generation show that SPATSG demonstrates superior semantic controllability, effectively reveals risk events across ADS, and maintains diversity and realism compared to baselines. This work offers a principled, interpretable approach for semantically controllable ADS testing and evaluation.