Can Attribution Predict Risk? From Multi-View Attribution to Planning Risk Signals in End-to-End Autonomous Driving

2026-05-07

AI Summary

This work addresses the opacity of end-to-end autonomous driving systems by introducing a hierarchical attribution framework that identifies visual regions critical to trajectory prediction through coarse-to-fine, multi-view region search. For the first time, it directly links multi-view attribution with planning risk prediction. The authors propose three generalizable attribution statistics—Attribution Entropy, intra-view spatial variance, and cross-camera Gini coefficient—to quantify the concentration and unevenness of model reliance on visual evidence. Experimental results demonstrate that the derived attribution signals significantly correlate with trajectory error (Spearman ρ = 0.30 ± 0.07) and achieve an AUROC of 0.77 ± 0.04 for collision detection, maintaining robust performance even in unseen scenarios.

Abstract

End-to-end autonomous driving models generate future trajectories from multi-view inputs, improving system integration but introducing opaque decisions and hard-to-localize risks. Existing methods either rely on auxiliary monitoring models or generate textual explanations, but are decoupled from the planning process and fail to reveal the visual evidence underlying trajectory generation. While attribution offers a direct alternative, planning differs from image classification by taking six-view camera images as input and predicting continuous multi-step trajectories, requiring attribution to capture both critical views and regions and their influence on outputs. Moreover, whether attribution maps can support risk identification remains underexplored. To address this, we propose a hierarchical attribution framework for end-to-end planning. Specifically, using L2 consistency with the original trajectory as the objective, we design a coarse-to-fine region attribution strategy that searches candidate regions across the full six-view input and refines attribution within them. We further extract three attribution statistics as predictive signals for planning risk, including attribution entropy to measure how concentrated the planner's reliance is over the joint visual space, within-camera spatial variance to characterize how spread out the attribution is within each view, and cross-camera Gini coefficient to quantify how unevenly attribution is distributed across the six cameras. Experiments on BridgeAD, UniAD, and GenAD show that these statistics correlate with planning risk, achieving Spearman correlations of $0.30 \pm 0.07$ with trajectory error and AUROC of $0.77 \pm 0.04$ for collision detection. The signal generalizes to held-out scenes with negligible degradation and remains stable under an alternative attribution baseline.

Problem

• attribution
• risk prediction
• end-to-end autonomous driving
• multi-view perception
• planning risk

Innovation

• attribution
• end-to-end autonomous driving
• planning risk
• multi-view cameras
• risk prediction