In Intelligent Cooperative Systems Laboratory, we are investigating and developing an “Intelligent Environment” (Environmental Robot). Environmental robot is quite different from humanoid robot (ex. ASIMO, HRP). Environmental robot has many variations of shape, structure and appearance. Most robots have quite simple functions and are embedded in a room. And they cooperate with other robots to support humans as if the whole room itself were a robot. In that sense, environmental robot doesn’t look like a “robot” but just a room or space. As a pioneer of the environmental robot, we have been developing “Robotic Room” since 1992. The latest version is RR3 (Robotic Room 3). In RR3 there are a ceiling mobile robot, container cases with intelligence, and a flower like robot. Please watch a near future living space surrounded by many un-robot-like robots.

We are developing "Sensing Room", which is a room that is equipped with massive sensors and computational devices. The room can monitor the occupants without constraints by the sensors. There are the following research topics on Sensing Room. 1) Sensor network software and accumulation system for massive and heterogeneous sensor data in room environment. 2) Wearable devices and wireless tiny device for behavior recognition and object location estimation. 3) Algorithm for understanding of daily human behavior patterns and anomaly detection from measured sensor data. 4) Framework for support at precise timing with appropriate devices based on recognized human behavior.

Because of the arrival of cheap and easy-to-use networked sensors in recent years, recognizing human action and analyzing human activity from sensors is of increasing interest. It is expected to grow advanced applications in area such as pervasive computing, security and human robot interaction, and fundamental issue on cognitive science and behavior sociology. Massive multi-modal sensor stream helps to promote a new region of artificial intelligence algorithms. We’ve explored a practical statistical algorithm for recognition and learning of human activity in the last decade. Recent research results include fast markerless motion capture from multi-view cameras, multi-label action annotation from motion captured data, multi-body detection-and-tracking with laser range finders, and human location estimation from micro waves.