Learning A Unified Risk Map for Autonomous Driving in Partially Observable Environments

2026-05-21

AI Summary

This work addresses the challenge of accurately assessing risk in occluded regions within partially observable environments, where existing autonomous driving methods often overestimate risk or produce inaccurate trajectory predictions. The authors propose a unified framework for risk map modeling and learning that, for the first time, integrates traffic-flow risk and collision risk into a joint spatiotemporal representation. To mitigate the scarcity of real-world occlusion interaction data, they introduce an adversarial occlusion scenario generation strategy based on diffusion models. This end-to-end framework enables fine-grained risk assessment and risk-aware planning, achieving significant improvements over current baselines on the Waymo Open Motion Dataset—specifically, increasing the minimum and average time-to-collision by 0.78× and 1.67×, respectively.

Abstract

Occlusion-aware prediction remains a critical challenge in autonomous driving due to the inherent uncertainty of unobserved regions. Existing approaches either overestimate risk based on reachable states or struggle to predict accurate trajectories under high occlusion uncertainty. To address these limitations, we propose a unified risk map modeling and learning framework for partially observable environments. Our method integrates traffic flow risk and collision risk through spatiotemporal modeling, enabling fine-grained assessment of occlusion-induced hazards. To address the scarcity of scenarios involving occluded interactions, we introduce a diffusion-based scenario generation framework that produces realistic yet adversarial scenarios. We integrate the modeling and learning of a unified risk map into a framework that supports risk-aware planning under partial observability. Experiments on the Waymo Open Motion Dataset show that our method significantly outperforms the state-of-the-art occlusion-aware baseline, improving minimum time-to-collision by 0.78 times and average time-to-collision by 1.67 times. The proposed framework offers a comprehensive and practical solution for risk-aware planning in partially observable environments.

Problem

• occlusion-aware prediction
• partially observable environments
• risk assessment
• autonomous driving
• uncertainty

Innovation

• unified risk map
• occlusion-aware prediction
• diffusion-based scenario generation
• spatiotemporal modeling
• risk-aware planning