cs.RONov 19, 2026

PAD: Policy-Aware Data Valuation for Autonomous Driving

Authors: Chengkai Xu, Yicheng Guo, Jiaqi Liu, Yiming Cui, Shiyu Fang, Peng Hang, Hao Zhang, Jian Sun

Organizations: College of Transportation, Tongji University · Department of Computer Science, UNC Chapel Hill · Department of Control Science and Engineering, Tongji University

Abstract

End-to-end autonomous driving increasingly depends on scaling training demonstrations, yet data growth does not guarantee proportional gains. However, additional data do not guarantee better policies: a demonstration that benefits one planner may be redundant or harmful to another. Existing curation methods typically infer data value from intrinsic sample properties or training-dependent proxies that do not measure downstream policy utility. To address this, we introduce Policy-Aware Data Valuation (PAD), a framework that assigns each candidate demonstration a utility specific to the target driving policy. PAD combines scene-level calibration feedback with projected policy gradients and a regularized approximation of local training geometry, producing scalable influence-based values for the policy. A controlled group-removal study establishes the directional fidelity of these values, while cross-planner analysis shows that the same candidate pool induces markedly different rankings across planner families. On DiffusionDrive, PAD improves EPDMS over real-only training by 3.53.5 points on NavTest and 5.05.0 points on NavHard. Policy-specific curation consistently outperforms real-only and full-pool training across diffusion-, regression-, and flow-matching-based planners while substantially reducing scene-level regressions. These results establish policy-aware data valuation as a basis for replacing indiscriminate data accumulation with targeted policy improvement, enabling more efficient and reliable scaling of autonomous-driving systems.

Figures & tables

Appendix figures & tables9 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Scaling-Aware Data Selection for End-to-End Autonomous Driving Systems

    Apr 9, 2026Tolga Dimlioglu, Nadine Chang, Maying Shen +2Training Data SelectionData-Efficient Learning

  2. FPD-Drive: Future Policy Distillation for End-to-End Autonomous Driving

    Oct 9, 2026Li-Heng Chen, Qifeng Chen, Chengwei Wu +4End-to-End Autonomous DrivingAutonomous Driving Planning