The Happy Robot team develops an open-hardware platform Happy Edu to lower entry barriers for RoboCup@Home and promote modular, extensible service robots. Our approach integrates multimodal perception, including speech interaction, open-vocabulary object recognition, and human-aware understanding.
Scientific achievements include a cost-effective 3D LiDAR–based re-identifiable human-following system that enables re-detection even after occlusion, ensuring robust tracking in domestic environments. We also proposed a foundation-model-driven planning framework for shelf display and disposal tasks, enabling adaptive decision-making without handcrafted rules. Additionally, we developed a robotic hand with four primitive motions for Bento assembly and a Fluorescent AR Marker method for automatic 6DoF pose annotation and estimation.
コメント