Generative Spatial Grounding in Free Space
A framework for restoring frame of reference in free-space human-robot communication.
This project explores how robots can communicate about locations in free space when there is no stable physical reference object.
The core idea is to move from selection-based grounding to generation-based grounding. Instead of only estimating a target point, the robot actively generates visual cues such as anchors or grids through augmented reality, reconstructing a shared frame of reference for spatial negotiation.
This direction connects robotics, XR, and spatial communication, and aims to support more precise and human-aligned interaction in open or ambiguous environments.