A Wearable IMU-Based Human-Robot Interaction Framework Configured with Large Language Models Büyük Dil Modelleri ile Yapilandirilan Giyilebilir IMU Tabanli ?Insan-Robot Etkileşimi Çerçevesi
34th Signal Processing and Communications Applications Conference, SIU 2026, İstanbul, Turkey, 7 - 10 July 2026, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/siu71813.2026.11636832
- City: İstanbul
- Country: Turkey
- Keywords: human-robot interation, IMU, LLM
- Middle East Technical University Affiliated: Yes
Abstract
In this study, a scalable and hardware-independent modular framework is proposed for IMU-based human-robot interaction. The architecture consists of a hardware abstraction layer for sensor integration, a calibration layer for sensor-segment alignment, a kinematic reconstruction layer for estimating human limb poses, and an interaction layer where human motions are interpreted as meaningful actions from the robot's perspective. This layered structure facilitates the integration of different wearable sensor systems while providing a consistent interface for higher-level interaction modules. In addition, the system aims to reduce setup complexity and improve adaptability by enabling high-level information - such as sensor placements and task descriptions - to be interpreted by large language models (LLM), allowing sensor mappings, kinematic model configurations, and calibration procedures to be generated automatically.