MomWorld: Momentum-Aware Latent World Model for Long-Horizon Autonomous Driving

2026-09-27

AI Summary

This study addresses the challenge in long-horizon autonomous driving planning where historical motion trends fail to propagate into the future, leading to momentum decay and unreliable near-term plans. To this end, we propose a Momentum-Aware Latent World Model that introduces a learnable momentum persistence mechanism, scene-conditioned updates, and an adaptive reset gate to jointly predict future configurations and momentum states. Furthermore, by integrating a MoFlow module with flow matching techniques, the method achieves trajectory-aligned scene evolution through few-step optimization. Evaluations on benchmarks such as NAVSIM demonstrate that the proposed model significantly enhances long-horizon planning consistency, reducing the average collision rate by 12.2% over baselines within a six-second horizon.

Abstract

Long-horizon planning enables autonomous vehicles to anticipate scene evolution and potential risks, supporting safe and stable decisions in complex interactions. However, existing methods struggle to propagate motion trends from observed history into the future. Long rollouts based on a single latent state may further attenuate useful dynamics, retain stale motion patterns, and disrupt reliable near-term plans. We introduce MomWorld, a momentum-aware latent world model for long-horizon planning. MomWorld extracts scene motion trends from historical-to-current observations and propagates latent momentum into future horizons, jointly predicting future configuration and momentum states. A learnable momentum persistence mechanism preserves stable trends, scene-conditioned momentum updates adapt future dynamics, and a scene-adaptive reset gate suppresses stale momentum under abrupt changes. We further propose MoFlow, a momentum-conditioned flow-matching module that refines a base trajectory to align with the predicted future scene evolution in only a few integration steps, with a horizon-aware residual fusion that preserves near-term planning stability while permitting stronger long-range corrections. Extensive experiments on NAVSIM, nuScenes and Bench2Drive demonstrate that MomWorld improves long-horizon planning consistency and reduces the average collision rate by 12.2% relative to MomAD over a 6-second planning horizon.

Problem

• long-horizon planning
• autonomous driving
• motion trend propagation
• latent world model
• stale motion patterns

Innovation

• Momentum-Aware Latent World Model
• Long-Horizon Planning
• Flow Matching
• Momentum Persistence Mechanism
• Autonomous Driving