Advanced Intelligent Systems Robotics and AI Lab

Research Theme

Ceiling-Climbing Robot

Mobile Module (MoMo) is a ceiling-Climbing robot. By moving along the ceiling, it can navigate the space without obstructing users.

Back to Intelligent Robotics

Research Focus

Comprehensive Sensing Using Multi-Agent Reinforcement Learning

This study aims to realize safe collaboration between humans and robots by developing a cooperative tracking system using multiple MoMo robots. For humans and robots to work together safely, it is essential to understand the surrounding environment with few blind spots and high accuracy. In dynamic environments, robots must continuously track and observe target objects or people while autonomously determining their positions and roles according to the situation. To address this challenge, this study applies multi-agent reinforcement learning to control a group of MoMo robots so that they can observe the target from multiple directions without losing coverage. Each robot is assigned a dynamic role, such as a front-view camera, rear-view camera, or overview camera. These roles are switched flexibly in response to the target's motion and changes in the environment. Through this approach, the system aims to achieve seamless and comprehensive sensing, enabling real-time, high-quality spatial understanding in complex human-robot collaborative environments.

High-Speed, Stable Movement of the Ceiling-Mounted Mobile Robot

This study proposes an improved structure for the Mobile Module (MoMo), a ceiling-mounted mobile robot, to achieve faster, more stable movement. MoMo moves while ceiling protrusions pass through rails mounted on its body, and springs press the wheels against the ceiling surface. However, in the conventional structure, friction and interference between the rails and protrusions made movement unstable. To address this issue, both the rail and protrusion structures were improved. The rail was redesigned as a two-layer structure, reducing joints and suppressing interference with the protrusions. In addition, the material was changed from aluminum to acrylic and from polycarbonate to acrylic to reduce weight and improve adaptability to height differences by leveraging their slight deformability during contact. Both ends of the protrusions were designed with circular arcs to improve step-passing performance. As a result, improvements in moving speed, maximum payload, and movement stability were confirmed. Future work will examine autonomous recovery from abnormal states, abnormality detection, and avoidance methods.