Latent World Models with Monotone Planning Costs for Image-Goal Navigation
2026-08-10 • Robotics
Robotics
AI summaryⓘ
The authors focus on improving image-goal navigation by creating a better way to predict and plan movements toward a target image. They measure planning cost using the similarity between predicted future states and the goal, but found that bad cost estimates can mislead planners. To fix this, they train a latent world model with special losses that help the model rank actions more reliably and handle multi-step predictions better. Their method works better than previous models on a navigation dataset and even succeeds on a real robot without extra training.
image-goal navigationlatent world modelcosine distanceCross-Entropy MethodDINO encoderautoregressive rollout lossMonotone Cost RankingInfoNCE losszero-shot transferrobot navigation
Authors
Amirhosein Chahe, Siwei Cai, Lifeng Zhou
Abstract
Image-goal navigation with latent world models requires not only accurate future prediction, but also a planning cost that reliably ranks candidate action sequences. We define the cost as the cosine distance between the predicted future embedding and the goal embedding, and show that poor cost ordering can mislead sampling-based planners such as Cross-Entropy Method (CEM). To address this, we propose a latent world model built on a frozen DINO-family encoder and train it with two complementary objectives. An autoregressive rollout loss reduces the gap between training and multi-step planning rollouts, while a Monotone Cost Ranking (MCR) loss directly encourages increasingly perturbed action sequences to receive higher planning costs. We also study InfoNCE-based action-contrastive training and find that temporal permutation negatives distort the latent geometry and degrade planning performance. On the GNM navigation dataset, our method outperforms Navigation World Models (NWM), DINO-WM, OmniVLA, and NoMaD, achieving state-of-the-art image-goal navigation performance while reducing orientation error by $2.7\times$ over the same-encoder DINO WM baseline. We also deploy the model zero-shot on a physical robot, where it follows goal-directed paths in unseen indoor and outdoor environments.