LiteNWM: Efficient Latent World Models for Onboard Visual Navigation in the Wild

Key points

• Direct visual navigation policies generate trajectories efficiently but do not explicitly evaluate their future consequences.1
• Generative navigation world models provide this foresight through visual rollouts, which are costly when evaluating multiple candidates.1
• In real-robot experiments in unseen indoor and outdoor environments, LiteNWM improves navigation success from 43.3% to 83.3% relative to NoMaD.1
• These results demonstrate that LiteNWM can be deployed for future-aware planning and closed-loop navigation on a physical robot.1

Sources (1)

• [1]LiteNWM: Efficient Latent World Models for Onboard Visual Navigation in the Wild
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 05:29 PM

  We present LiteNWM, a latent navigation world model that shares visual encoding across candidates and jointly predicts their action-conditioned future representations at multiple horizons, while a learned scorer uses these predictions to select trajectories.

  Direct visual navigation policies generate trajectories efficiently but do not explicitly evaluate their future consequences.

Extractive summary: sentences quoted from the sources.