OWMDrive: Causality-Aware End-to-End Autonomous Driving via 4D Occupancy World Model

2026-06-29

AI Summary

Existing end-to-end autonomous driving approaches struggle to generate safe and stable trajectories in occluded or highly uncertain scenarios due to their lack of explicit future modeling of traffic dynamics. This work proposes the first generative end-to-end framework that integrates a causal-aware 4D occupancy world model with a diffusion-based planner. By leveraging multi-step 3D occupancy predictions as conditional priors, the method explicitly captures spatiotemporal causal dependencies among traffic participants, enabling forward-looking trajectory planning. Evaluated under partially observable and challenging driving conditions, the approach significantly enhances planning safety and reliability, outperforming current end-to-end baselines.

Abstract

Autonomous driving systems are steadily moving toward end-to-end paradigms to mitigate the limited adaptability of rule-based pipelines in complex traffic environments. However, most existing learning-based methods still make decisions from static representations of the current scene, without explicit future rollouts or modeling of the temporal causal dynamics in traffic interactions. This limitation often results in unstable or overly conservative planning under high-uncertainty conditions, such as occlusions and unexpected events. To overcome these challenges, we introduce OWMDrive, a generative end-to-end driving framework built upon an Occupancy World Model for multi-step 3D occupancy forecasting, which serves as a conditional prior to guide diffusion-based planning. Conditioned on both current observations and predicted future states, the planner iteratively refines trajectory candidates to generate a reinforced driving trajectory. By explicitly modeling scene evolution over future horizons, OWMDrive captures key spatiotemporal causal dependencies, which leads to more foresighted and robust trajectory generation. Extensive experiments demonstrate that OWMDrive significantly improves planning reliability and safety, especially in challenging and partially observable driving scenarios.

Problem

• autonomous driving
• causality
• temporal dynamics
• occupancy forecasting
• end-to-end planning

Innovation

• Occupancy World Model
• Causal Dynamics
• End-to-End Autonomous Driving
• Diffusion-based Planning
• 4D Scene Forecasting