GeoDrive: 3D Geometry-Informed Driving World Model with Precise Action Control

TL;DR

GeoDrive introduces a 3D geometry-informed world model for autonomous driving that improves spatial understanding, action control, and scene editing, addressing previous deficiencies in geometric consistency and occlusion handling.

Contribution

The paper presents GeoDrive, a novel driving world model that explicitly incorporates 3D geometry and dynamic editing to enhance safety, accuracy, and scene manipulation in autonomous driving.

Findings

Outperforms existing models in action accuracy and 3D spatial awareness

Enables realistic and adaptable scene modeling for safer driving

Supports interactive scene editing and trajectory control

Abstract

Recent advancements in world models have revolutionized dynamic environment simulation, allowing systems to foresee future states and assess potential actions. In autonomous driving, these capabilities help vehicles anticipate the behavior of other road users, perform risk-aware planning, accelerate training in simulation, and adapt to novel scenarios, thereby enhancing safety and reliability. Current approaches exhibit deficiencies in maintaining robust 3D geometric consistency or accumulating artifacts during occlusion handling, both critical for reliable safety assessment in autonomous navigation tasks. To address this, we introduce GeoDrive, which explicitly integrates robust 3D geometry conditions into driving world models to enhance spatial understanding and action controllability. Specifically, we first extract a 3D representation from the input frame and then obtain its 2D rendering based on the user-specified ego-car trajectory. To enable dynamic modeling, we propose a dynamic editing module during training to enhance the renderings by editing the positions of the vehicles. Extensive experiments demonstrate that our method significantly outperforms existing models in both action accuracy and 3D spatial awareness, leading to more realistic, adaptable, and reliable scene modeling for safer autonomous driving. Additionally, our model can generalize to novel trajectories and offers interactive scene editing capabilities, such as object editing and object trajectory control.