DrivingWorld: Constructing World Model for Autonomous Driving via Video GPT

TL;DR

DrivingWorld introduces a novel GPT-style model with spatial-temporal fusion mechanisms for autonomous driving, enabling high-fidelity, long-duration video generation and improved control over future scene prediction.

Contribution

The paper presents a new spatial-temporal fusion GPT-style model with strategies for better generalization and control, significantly surpassing prior models in video quality and duration.

Findings

Generates high-quality driving videos over 40 seconds long

Achieves over twice the duration of previous state-of-the-art models

Outperforms prior methods in visual fidelity and controllability

Abstract

Recent successes in autoregressive (AR) generation models, such as the GPT series in natural language processing, have motivated efforts to replicate this success in visual tasks. Some works attempt to extend this approach to autonomous driving by building video-based world models capable of generating realistic future video sequences and predicting ego states. However, prior works tend to produce unsatisfactory results, as the classic GPT framework is designed to handle 1D contextual information, such as text, and lacks the inherent ability to model the spatial and temporal dynamics essential for video generation. In this paper, we present DrivingWorld, a GPT-style world model for autonomous driving, featuring several spatial-temporal fusion mechanisms. This design enables effective modeling of both spatial and temporal dynamics, facilitating high-fidelity, long-duration video generation. Specifically, we propose a next-state prediction strategy to model temporal coherence between consecutive frames and apply a next-token prediction strategy to capture spatial information within each frame. To further enhance generalization ability, we propose a novel masking strategy and reweighting strategy for token prediction to mitigate long-term drifting issues and enable precise control. Our work demonstrates the ability to produce high-fidelity and consistent video clips of over 40 seconds in duration, which is over 2 times longer than state-of-the-art driving world models. Experiments show that, in contrast to prior works, our method achieves superior visual quality and significantly more accurate controllable future video generation. Our code is available at https://github.com/YvanYin/DrivingWorld.