ProDrive: Proactive Planning for Autonomous Driving via Ego-Environment Co-Evolution

2026-04-28

AI Summary

This work addresses the limitations of conventional end-to-end autonomous driving planners, which rely solely on current observations and often yield myopic decisions that compromise safety. The authors propose ProDrive, a novel framework that establishes the first bidirectional coupling between a trajectory planner and a world model. Specifically, planning-aware ego-vehicle tokens inject candidate trajectory information into bird’s-eye-view (BEV) environmental predictions, enabling the joint end-to-end optimization—via gradient flow—of a query-centric trajectory planner and the world model. This architecture supports parallel evaluation of multiple candidate trajectories in terms of their impact on future scene evolution. Experiments on NAVSIM v1 demonstrate that ProDrive significantly outperforms strong baselines in both safety and planning efficiency, while ablation studies confirm the effectiveness of the proposed ego-environment co-evolution mechanism.

Abstract

End-to-end autonomous driving planners typically generate trajectories from current observations alone. However, real-world driving is highly dynamic, and such reactive planning cannot anticipate future scene evolution, often leading to myopic decisions and safety-critical failures. We propose ProDrive, a world-model-based proactive planning framework that enables ego-environment co-evolution for autonomous driving. ProDrive jointly trains a query-centric trajectory planner and a bird's-eye-view (BEV) world model end-to-end: the planner generates diverse candidate trajectories and planning-aware ego tokens, while the world model predicts future scene evolution conditioned on them. By injecting planner features into the world model and evaluating all candidates in parallel, ProDrive preserves end-to-end gradient flow and allows future outcome assessment to directly shape planning. This bidirectional coupling enables proactive planning beyond current-observation-driven decision-making. Experiments on NAVSIM v1 show that ProDrive outperforms strong baselines in both safety and planning efficiency, while ablations validate the effectiveness of the proposed ego-environment coupling design.

Problem

• autonomous driving
• proactive planning
• trajectory planning
• scene evolution
• end-to-end planning

Innovation

• proactive planning
• ego-environment co-evolution
• world model
• end-to-end autonomous driving
• BEV prediction