A paper by the research team of Professor Hyung-Gi Cho of Jeonbuk National University (JBNU) (College of Engineering, Division of Electronic Engineering, AI Robotics Lab) was finally accepted to the IEEE International Conference on Robotics and Automation (ICRA 2026), the world’s most prestigious conference in robotics.

ICRA is regarded as the most influential international conference in the fields of robotics and autonomous systems. The conference is organized by the IEEE Robotics and Automation Society and will be held in Vienna, Austria from June 1 to June 5, 2026.

The accepted paper is titled “SC-VLMaps: Depth-Free Visual–Language Mapping via Scene Coordinate Regression.” The research team proposed a new method that constructs a 3D spatial map and linguistic semantic information simultaneously using only a monocular RGB camera, without the expensive depth sensors required by existing Visual–Language Map approaches.

The study utilizes an AI model based on Scene Coordinate Regression (SCR) to directly predict 3D coordinates from images, and combines these predictions with the semantic representations of large language models, fundamentally overcoming sensor constraints.

This enables robots to perform spatial understanding and exploration based on natural-language queries such as "Where is the table?" and "Find the area with windows."

In particular, the result is significant because it demonstrates that high-level spatial intelligence can be achieved even in low-cost sensor environments. It is regarded as international recognition of core technologies in Spatial AI and Physical AI.

The research was conducted as a global collaborative study with an overseas robotics company and was carried out with support from the National Research Foundation of Korea’s Excellent Early-Career Research Program (RS-2024-00346415).

Professor Cho said, "This achievement is a case that demonstrates precise 3D spatial understanding is possible even in low-cost sensor environments," and added, "We will expand the research into autonomous behaviors and human–robot interaction on various robot platforms in the future."