HorizonDrive: Self-Corrective Autoregressive World Model for Long-horizon Driving Simulation
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866917522082627584 |
|---|---|
| author | Zhang, Conglang Zhan, Yifan Wang, Qingjie Ouyang, Zhanpeng Li, Yu Yang, Zihao Guo, Xiaoyang Ren, Weiqiang Zhang, Qian Dong, Zhen Zheng, Yinqiang Yin, Wei Chen, Zhengqing |
| author_facet | Zhang, Conglang Zhan, Yifan Wang, Qingjie Ouyang, Zhanpeng Li, Yu Yang, Zihao Guo, Xiaoyang Ren, Weiqiang Zhang, Qian Dong, Zhen Zheng, Yinqiang Yin, Wei Chen, Zhengqing |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_11596 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | HorizonDrive: Self-Corrective Autoregressive World Model for Long-horizon Driving Simulation Zhang, Conglang Zhan, Yifan Wang, Qingjie Ouyang, Zhanpeng Li, Yu Yang, Zihao Guo, Xiaoyang Ren, Weiqiang Zhang, Qian Dong, Zhen Zheng, Yinqiang Yin, Wei Chen, Zhengqing Computer Vision and Pattern Recognition 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. |
| title | HorizonDrive: Self-Corrective Autoregressive World Model for Long-horizon Driving Simulation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.11596 |