Integrated Sensing: Virtual Sensor-Actuator Arrays

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

This project concerns the interface between application software and the real-time data in large-scale arrays of wireless sensors and actuators. The objective is to develop the theory and algorithms for creating a virtual sensor-actuator array (VSAA) that will make it possible for application algorithms to avoid dealing with the continually changing features of physical sensor-actuator arrays. Using the VSAA, application-specific software can be written as if the sensing and actuating devices are located as assumed in the application design model, and that information is received with appropriate time stamps, in addition to being filtered and integrated in a near optimal manner. This project will build on recent research in intelligent data integration for distributed systems. Raw data will be delivered and manipulated as soft decisions in the form of probability distributions. Information from multiple sources will be integrated using novel algorithms based on graphical representations of the information relationships among the sources, and information-theoretic techniques will be developed for sharing and moving the computations among the nodes so that performance does not degrade significantly in the face of intermittent communication and node failures. The algorithms developed in this project will be evaluated for intelligent building control applications using in a wireless sensor-actuator array in the Intelligent Workplace, an existing experimental facility at Carnegie Mellon University.