NavThinker: Action-Conditioned World Models for Coupled Prediction and Planning in Social Navigation

Tianshuai Hu, Zeying Gong, Lingdong Kong, XiaoDong Mei, Yiyi Ding, Qi Zeng, Ao Liang, Rong Li, Yangyi Zhong, Junwei Liang · 2026

Read NavThinker: Action-Conditioned World Models for Coupled Prediction and Planning in Social Navigation by Tianshuai Hu, Zeying Gong, Lingdong Kong, XiaoDong Mei, Yiyi Ding, Qi Zeng, Ao Liang, Rong Li, Yangyi Zhong, Junwei Liang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-15359.txt.

Description

Hu and co-authors present NavThinker, a framework that couples prediction and planning for robots moving among people. Safe navigation among people needs foresight about how paths interact. An action-conditioned world model predicts future scene depth and human motion for each candidate action. Those forecasts feed the policy as extra features and shape its social rewards. The system set state-of-the-art success on Social-HM3D tests. It transferred to a new benchmark with no retraining and ran on a real Unitree Go2 robot.

Catalog record

CREATORTianshuai Hu, Zeying Gong, Lingdong Kong, XiaoDong Mei, Yiyi Ding, Qi Zeng, Ao Liang, Rong Li, Yangyi Zhong, Junwei Liang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34923
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.15359

Browse the full library · About the Tower of Babel