Momentum aware model improves long term planning for self driving cars
MomWorld: Momentum-Aware Latent World Model for Long-Horizon Autonomous Driving
Robotics
Summary
Planning far into the future helps self-driving cars make safer choices by anticipating what other cars and pedestrians will do. The authors found that current models have trouble keeping track of ongoing movement trends, which makes their predictions less reliable over time. They created MomWorld, a model that remembers momentum from past observations and adjusts predictions to stay accurate even during sudden changes. Tests on real and simulated driving data showed that MomWorld reduces crashes and improves planning over six seconds ahead.
What this means in practice
- •For autonomous vehicle developers: Generate more reliable long-term motion predictions to improve safety and responsiveness in self-driving car planning systems.
- •For robotics navigation engineers: Enhance momentum-aware trajectory forecasting in mobile robots for safer navigation in dynamic environments.
Authors
Ziying Song, Shengkai Zhang, Lei Yang, Haozhuang Chi, Yuchen Liu, Jiangtao Su, Lin Liu, Ziyang Liu, Chen Lv
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.