GeoWorldAD: Geometry World Action Model for Autonomous Driving

2026-07-19

AI Summary

This work addresses the challenge of balancing safety and efficiency in autonomous driving within dynamic 3D environments, where existing vision-based action models often lack explicit geometric constraints and the ability to predict future scene evolution. To overcome these limitations, the authors propose an ego-centric 3D geometric world model that, for the first time, incorporates explicit 3D geometric grounding and short-horizon future geometric representations. A multi-scale geometric aggregation mechanism is introduced, leveraging a vision Transformer architecture to fuse current observations with implicit future geometric features, enabling end-to-end trainable, iterative trajectory refinement. Evaluated on the NAVSIM v1 and v2 benchmarks, the method achieves state-of-the-art performance, demonstrating that explicit 3D geometric modeling effectively enhances both decision-making safety and driving efficiency.

Abstract

Autonomous driving requires both safe and efficient planning decisions in dynamic 3D environments. Although recent Vision/Video-Action models learn policies directly from visual observations and scale well with advances in vision transformers and large-scale training data, they often lack explicit geometric grounding and future-aware spatial guidance, limiting their ability to balance collision avoidance and driving progress. In this work, we propose GeoWorldAD, a geometry world action model that grounds trajectory planning in ego-aligned 3D space and anticipates short-horizon scene evolution with latent future geometry tokens. Present geometry provides essential spatial constraints for safe planning, while future geometry reveals how surrounding agents and ego-centric free space may evolve, reducing overly conservative decisions without sacrificing safety. To efficiently exploit these geometric cues, GeoWorldAD progressively aggregates multi-scale present geometry and latent future geometry through iterative trajectory refinement. Experiments on NAVSIM v1 and v2 demonstrate state-of-the-art performance, highlighting the effectiveness of explicit 3D geometry grounding and future geometry world modeling for safe and efficient autonomous driving.

Problem

• autonomous driving
• 3D geometry grounding
• future-aware planning
• collision avoidance
• trajectory planning

Innovation

• 3D geometry grounding
• future geometry anticipation
• ego-aligned trajectory planning
• latent future tokens
• iterative trajectory refinement