Researchers to Improve Efficiency, Reliability of Networked Systems
From taking soil moisture measurements around the country to using sensing for surveillance, many modern technologies rely on networked systems.
But too often, these systems are not as efficient or reliable as they could be. That is largely because networked systems are informationally decentralized, comprise many nodes carrying disparate information and are subject to constraints on energy, data storage and computational capabilities.
Researchers in the University of Illinois Information Trust Institute and Coordinated Science Laboratory, along with investigators at the University of Michigan at Ann Arbor, aim to address this problem by developing a general theoretic framework and tools to help optimize these sensing systems. They received a five-year, $2.5 million grant, titled CIF: Large: Collaborative Research: Controlled Sensing, and Distributed Signal Processing and Decision Making in Networked Systems, funded by the National Science Foundation.
People are building these sensing systems with a large number of nodes, but without a theory to optimize them, said professor Venu Veeravalli, Illinois's principal investigator and professor of electrical and computer engineering. The sensors have to communicate, they have to coordinate and they have to operate in a limited resource environment, which means they have to be efficient.
Researchers will study the role of information in sensing, signal processing and decision making for networked systems under various architectures, in both controlled and distributed sensing. They will also work to understand the coordination of networked systems and develop novel algorithms to enhance the functioning of these systems. In addition to three investigators at Michigan, the team also includes professors Tamer Başar, professor of electrical and computer engineering, and Angelia Nedich, assistant professor of industrial & enterprise systems engineering.
To maximize efficiency, researchers will develop new event-driven sensing techniques. Today's sensors turn on and off arbitrarily to conserve energy. They may be on just 10 percent of the time, meaning they could miss significant events. The investigators aim to develop a smart system that functions as though it were on 100 percent of the time, even if it's in wake mode for only a fraction of that time.
Sensors would evaluate local events and decide whether they are important, Başar said. They would only activate themselves if there's information that's worth collecting and transmitting.
Data storage is also an important component of the project. Sensors cannot store a significant amount of data, so they either eliminate it or transmit it. Researchers will develop intelligent storage, which helps identify what should be discarded and what should be transmitted.
The team will evaluate their findings using a soil moisture monitoring test bed at the University of Michigan. In addition, the research could apply to fire and rescue operations (sensors collecting data on heat sources) and any number of atmospheric and environmental operations, from the early detection of earthquakes to climate change to monitoring the impact of oil spills on oceanic life.
Veeravalli said: As we rely more and more on these technologies to further our understanding of the natural world and other systems, we need to use sensors in the most effective and efficient way.
January 3, 2012
by Kim Gudeman, Coordinated Scinece Lab