arXiv cs.AI· Ruiguo Zhong, Xiaolong Chen, Benshan Ma, Lang Zhang, Mingyue Feng, Yaonong Wang, Pei Liu, Jun Ma·· 12 小时前AI 评分31
DA-WAM：面向驾驶世界模型的决策对齐未来潜变量
DA-WAM: Decision-Aligned Future Latents for Driving World Models
AI 导读
DA-WAM 将 JEPA 与候选级轨迹打分结合，为每条候选轨迹生成独立未来潜变量，并用多头打分器估计规划因素与整体效用。该方法在 NAVSIM-v1、NAVSIM-v2 和 Bench2Drive 上取得基于摄像头的学习型规划器中的 SOTA 表现。针对几何相似但安全结果不同的轨迹，安全关键后训练与 hard negatives 进一步提升了区分能力。
正文 · 原文
Abstract:World models for autonomous driving should anticipate the consequences that matter for driving decisions rather than reconstruct every detail of future scenes. However, predictive representations alone are not guaranteed to distinguish candidate actions that lead to different planning this http URL propose DA-WAM, a decision-aligned world action model that couples a Joint-Embedding Predictive Architecture (JEPA) with candidate-level trajectory scoring. During joint training, the JEPA objective grounds future representations in observed scene dynamics, while the planning objectives encourage these representations to capture distinctions relevant to safety and utility. An action-conditioned predictor generates a distinct future latent for each candidate trajectory, and a multi-head trajectory scorer uses the corresponding latent to estimate planning factors and overall utility. In addition, safety-critical post-training with hard negatives further improves discrimination between geometrically similar trajectories with different safety outcomes. DA-WAM achieves state-of-the-art performance among learned camera-based planners on NAVSIM-v1, NAVSIM-v2, and Bench2Drive. Ablation studies demonstrate the effectiveness of each component in driving decision-making.
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.19085 [cs.RO] |
| (or arXiv:2608.19085v3 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2608.19085 arXiv-issued DOI via DataCite |
Submission history
From: Ruiguo Zhong [view email]
[v1]
Wed, 19 Aug 2026 16:33:02 UTC (2,743 KB)
[v2]
Thu, 20 Aug 2026 06:46:02 UTC (2,741 KB)
[v3]
Sun, 18 Oct 2026 12:20:37 UTC (1,925 KB)
来源：arXiv cs.AI · arxiv.org