image: Wei Ye: lead researcher in development of new sensornet communication protocol. view more
Credit: USC Information Sciences Institute
A new communication protocol for wireless sensor networks just released by the Viterbi School's Information Sciences Institute is the most efficient yet with more than a tenfold improvement on previous versions.
Sensor networks, or ÒsensornetsÓ are an emerging way to monitor inaccessible and unwired places. They depend on placing numerous sensor units across a wide area. The units communicate with each other, and send the information they gather at intervals to human operators.
In wilderness parks, for example, such networks are used to monitor activity by wildlife. Sensornets are also being explored for industrial applications Ñ in oilfield monitoring and managment, for example. Ordinary wireless methods, such as WiFi wonÕt work for this purpose.
The units are battery powered, so minimum power consumption is critical Ñ but at the same time, continuing coverage is essential.
The activities of the units are orchestrated by special operating rules called Media Access Control (MAC) protocols. More than three years of ISI research Ð supported by the National Science Foundation, Intel and other funders Ð produced a new protocol, SCP-MAC, which produced a dramatic improvement in energy efficiency.
The protocol combines two techniques: Ôlow power listeningÓ in which units switch on for only very brief periods; and Òscheduled channel pollingÓ which synchronizes and schedules the listening.
"The basic approach of SCP-MAC is to let units alternate periods of sleeping with very brief periods of listening, as shown in the figure," says Wei Ye, an ISI research scientist. "Such a sleep pattern is found on birds, who need to keep vigilance while sleeping. To minimize the listening cost, SCP-MAC utilizes 'low-power listening,' which detects channel activity very quickly.
"It further reduces the transmission cost by synchronizing the listening schedules of nodes, so that a unit can wake up its neighbors by transmitting a short tone."
Previous protocols required individual units to be active for approximately 2-3 per cent of monitoring time- that is, active about 29-45 minutes of every day of sensornet activity. SCP-MAC reduced the monitoring time to less than two minutes each day.
The system was developed by Ye, working with project leader John Heidemann and programmer Fabio Luis Silva in the ISI Laboratory for Embedded Networked Sensor Experimentation.
The trio gave a final presentation on the research in November at the Proceedings of the Fourth ACM SenSys Conference in Boulder Colorado.
In February, the group made the new protocol, written in the TinyOS operating system optimized for Mica2 individual sensor modules, available for download by sensornet users and developers.
The download url is: http://www.isi.edu/ilense/software/ scpmac/