Publications

Learning Category-level Last-meter Navigation from RGB Demonstrations of a Single-instance

Tzu-Hsien Lee, Fidan Mahmudova, Karthik Desingh
arXiv preprint arXiv:2512.11173, 2025.

Motivation. Achieving precise positioning of a mobile manipulator's base is essential for successful manipulation actions. However, most RGB-based navigation systems only guarantee coarse, meter-level accuracy.

Method. We introduce an object-centric imitation learning framework for last-meter navigation, enabling a quadruped mobile manipulator to achieve manipulation-ready positioning using only RGB observations. Trained on real-world data from a single object instance, the system generalizes to unseen object instances across diverse environments with challenging lighting and background conditions.