Generating Realistic Safety-Critical Scenarios for Vehicle-Pedestrian Interactions

2026-05-16

AI Summary

This work addresses the scarcity of high-risk vehicle-pedestrian interaction scenarios in real-world data and the inability of existing simulators to generate safety-critical cases with realistic behaviors. To overcome this, the authors propose a three-stage framework that combines pretraining on real-world data with online reinforcement learning in CARLA to scalably produce high-fidelity interactive scenarios. They introduce a novel multi-agent state-space Transformer-enhanced DDPG (MA-SST-DDPG) algorithm that integrates data-driven learning with adaptive simulation, enabling the generation of hazardous behaviors that are distributionally equivalent to real data and indistinguishable by human evaluators. The resulting VPSCI dataset comprises over 198,000 high-resolution interaction clips, achieving state-of-the-art trajectory prediction accuracy (ADE=0.072m, FDE=0.142m) and passing both Turing tests and statistical distributional equivalence checks.

Abstract

Automated driving system deployment requires rigorous validation across safety-critical vehicle-pedestrian interactions, yet real-world datasets rarely capture high-risk scenarios while simulation platforms lack realistic behavior. In response, this study proposes a three-stage framework that combines real-world grounding with adaptive simulation to generate behaviorally realistic safety-critical scenarios at scale. Stage 1 pre-trains multi-agent state-space Transformer-enhanced DDPG (MA-SST-DDPG) agents on real-world safety-critical data to learn human-like interactive evasive behaviors through data-driven learning. Stage 2 deploys pre-trained multi-agents in CARLA for online reinforcement learning to generalize across diverse scenarios, integrating real-world knowledge with simulation experience to produce a refined MA-SST-DDPG model. Stage 3 uses CARLA with the refined model to generate over 198,000 high-resolution interaction episodes from eight intersection scenarios, culminating in the Vehicle-Pedestrian Safety-Critical Interaction (VPSCI) dataset. The Refined MA-SST-DDPG model outperformed baseline methods in reproducing realistic evasive behaviors, achieving the lowest trajectory errors (ADE = 0.072 m, FDE = 0.142 m). Statistical comparison confirmed distributional equivalence between the generated and real-world data in both conflict severity and behavioral response. A Turing test confirmed that the three-stage framework generated evasive behaviors were indistinguishable from real-world interactions. These results demonstrate the framework's effectiveness in producing high-fidelity safety-critical data, offering valuable sources for the development of ADS and simulation-based safety evaluations.

Problem

• safety-critical scenarios
• vehicle-pedestrian interactions
• realistic behavior
• automated driving systems
• simulation validation

Innovation

• MA-SST-DDPG
• safety-critical scenario generation
• vehicle-pedestrian interaction
• realistic behavior modeling
• CARLA simulation