Ambient Intelligence for Better Buildings
Sensor networks have the potential to allow the development of truly intelligent buildings that improve productivity, efficiency, safety, and security. To be practical, such networks must be efficient, scalable to very large spaces, and economical to manufacture, install and maintain. One answer is networks of passive infrared motion detectors. Sensors could be manufactured onto building infrastructure elements such as light fixtures. We are developing technology that enables cost-effective networks to recognize, predict, and index human activity in building-scale environments.
Background & Objective: The falling cost of sensors, microprocessors, and radios combined with the rising interest in safe, secure, and efficient buildings leads us to believe that holistic building systems are the future of building management. Many groups across the globe are focused on the problem of better packet routing, better sensing modalities, and lower-power computing. At MERL we are leveraging that work by focusing on the perceptual question: what is it possible to sense with these systems, what are the fundamental structures of human activity at the building scale, what can we learn, and what benefit can we gain? We have assembled a host of experiments and demonstrations that illustrate the usefulness of these systems for security, efficiency, safety and situational awareness within buildings.
Technical Discussion: These results rest on the technical innovations of several other projects. Ultra-low power, easy to manufacture and easy to install, the Reduced Operating Cost Sensors (ROCkS) platform has enabled a host of experiments by allowing us to collect data from large, multi thousand square meter installations, over year-scale spans of time to assemble a unique dataset about human behavior at the building-scale. The Scalable Activity Recognition for Sensor Networks (SARSEN) project has focused on finding structures in that data, and probabilistic models to detect and analyze those structures. The Integrated Event Recognition project has focused using sensor networks to help solve problems that are hard given only classical sensor modalities.
Future Direction: We are pursuing several technical and business directions for the project. Technically, we are pursuing more detailed models of crowds, further refinements and validation of the ROCkS platform, and deeper interaction between sensor networks and classical modalities.
Contacts:
Christopher R. Wren
Yuri Ivanov
| Technical Reports: | |
| Visualizing the History of Living Spaces | |
| Tracking People in Mixed Modality Systems | |
| Worse is Better for Ambient Sensing | |
| Automatic Pan-Tilt-Zoom Calibration in the Presence of Hybrid Sensor Networks | |
Technology Areas:
Sensor and Data Systems
Computer Vision
Modification Date: July 1, 2008
