The goal of this research project is to build an application infrastructure and a suite of embedded system software that will enable the rapid development of software control and monitoring applications, such as heating, ventilation, and air conditioning systems (HVAC), on low-power, lossy wireless sensor-actuator networks. Such user-preference driven control applications require the ability to adapt to noise, loss, and failures in a large-scale distributed network, and demand a high degree of automation in acquiring user preferences, making control decisions and disseminating those decisions to the appropriate actuators.
The Sensor Control System (SCS) being developed provides a high-level platform that allows users to express control application requirements at the granularity of the entire network, enabling users to focus on the requirements of their deployments by abstracting away the low-level sensor network details. SCS combines techniques from machine learning and statistics with high-level abstractions inspired by work in software systems and databases, allowing for mathematically-sound decision making despite loss and uncertainty that is inherent in such systems. The project will demonstrate the approach on a highly instrumented office environment. This research has the potential to impact industry through the development of a new class of wireless control systems, and academia through the infusion of cross-disciplinary ideas into a variety of sub-fields of EECS. The datasets, code and the sensor-actuator deployments developed during the course of this research will be used in a range of educational initiatives.
