Project page
HorizonDrive Film mode preview
Preparing film Loading video scenes

Long-horizon driving world model

HorizonDrive

Self-Corrective Autoregressive World Model for Long-horizon Driving Simulation

Abstract

HorizonDrive keeps autoregressive driving simulation reliable beyond short clips.

Our key insight is to make the teacher rollout-capable, ensuring reliable supervision from its own AR rollouts.
Controllable Driving scene generation
Long-Horizon Stable generation quality
Interactive AR rollout
Scalable Diverse driving scenes
Closed-Loop Simulation ready
No Explicit 3D representations

Method Overview

We first train a conditional driving world model, then improve its autoregressive stability through scheduled rollout recovery, and finally distill long-horizon teacher rollouts into a few-step, short-chunk student via teacher-rollout DMD.

Overview of the HorizonDrive method pipeline. A three-stage method diagram showing conditional world model training, scheduled rollout recovery, and teacher rollout DMD. Stage1: Conditional Driving WM (Sec. 4.1) Stage2: Scheduled Rollout Recovery (Sec. 4.2 ) Stage3: Teacher Rollout DMD (Sec. 4.3) HD Map 3D Bboxs Action Conditional Driving WM Short Horizon ... Base Model Data Preparation Rollout ... Base Model Rollout Prediction Train degraded history clean target with noise Rollout-capable Base Model Rollout Teacher Teacher rollout DMD Student Long Horizon CFG truncation threshold 1000 300 no CFG CFG training iterations ... 1st chunk 2nd chunk 3rd chunk 4th chunk

Results

20-Second AR Results on Nuscenes

2x playback

Each model rollout generates 1 s of future video.

Results

30-Second AR Results on Self-Collected Dataset

2x playback

Each model rollout generates 1 s of future video.

Results

Minute-Level AR Video Generation

3x playback

Each model rollout generates 1 s of future video.

Simulation

Closed-Loop Driving Simulation

Each model rollout generates 1 s of future video.

HorizonDrive

Thank you for your interest in our work.

See more details at