让驾驶模型自修正，实现分钟级实时长时模拟。
HorizonDrive: Self-Corrective Autoregressive World Model for Long-horizon Driving Simulation
- 用预测误差恢复训练使教师模型抗漂移，支持长期自回归推演。
- 在nuScenes上FID降低52%，轨迹误差减少21%，超越现有长时流式基线。
- 适合需要长时间高精度驾驶仿真的自动驾驶研发团队使用。
封闭环路驾驶仿真需实现实时交互而非仅短片段离线生成，推动当前驾驶世界模型向自回归（AR）推演发展。现有AR蒸馏方法通常依赖帧丢失或学生端退化训练，前者因车辆快速运动和场景剧烈变化导致迁移效果差，后者受限于教师单次输出长度，监督范围有限。本文提出：能否通过自回归推演扩展教师模型，在内存可控条件下提供无界监督？核心难点在于标准教师在自身预测下会漂移，污染监督信号。我们的关键洞察是让教师具备自回归推演能力，确保其推演结果可靠。为此提出HorizonDrive，一种反漂移的训练与蒸馏框架。首先，计划性推演恢复（SRR）训练基础模型从预测污染的历史中重建真实未来片段，使教师在长时推演中保持稳定；其次，利用可推演教师通过自回归推演提供长时程分布匹配监督，同时短窗口学生通过教师推演动态匹配（TRD）对齐，实现高效实时部署。HorizonDrive原生支持分钟级自回归推演且内存可控；在nuScenes数据集上，相比最强的长时流式基线，FID降低52%，FVD降低37%，平均轨迹误差（ARE）下降21%，动态时间规整（DTW）降低9%，同时保持与单次生成驾驶视频模型相当的性能。
原文摘要 · Abstract (English)
Closed-loop driving simulation requires real-time interaction beyond short offline clips, pushing current driving world models toward autoregressive (AR) rollout. Existing AR distillation approaches typically rely on frame sinks or student-side degradation training. The former transfers poorly to driving due to fast ego-motion and rapid scene changes, while the latter remains bounded by the teacher's single-pass output length and thus provides only a limited supervision horizon. A natural question is: can the teacher itself be extended via AR rollout to provide unbounded-horizon supervision at bounded memory cost? The key difficulty is that a standard teacher drifts under its own predictions, contaminating the supervision it provides. Our key insight is to make the teacher rollout-capable, ensuring reliable supervision from its own AR rollouts. This is instantiated as HorizonDrive, an anti-drifting training-and-distillation framework for AR driving simulation. First, scheduled rollout recovery (SRR) trains the base model to reconstruct ground-truth future clips from prediction-corrupted histories, yielding a teacher that remains stable across long AR rollouts. Second, the rollout-capable teacher is extended via AR rollout, providing long-horizon distribution-matching supervision under bounded memory, while a short-window student aligns to it with teacher rollout DMD (TRD) for efficient real-time deployment. HorizonDrive natively supports minute-scale AR rollout under bounded memory; on nuScenes, HorizonDrive reduces FID by 52% and FVD by 37%, and lowers ARE and DTW by 21% and 9% relative to the strongest long-horizon streaming baselines, while remaining competitive with single-pass driving video generators.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。