Energy efficient sensor networks using analog signal processing

 

Wide-scale deployment of wireless sensor networks for envisioned applications in surveillance and monitoring has been inhibited primarily by the lack of ability to last for long durations on small power sources, such as batteries and energy-harvesting systems. Any effort to reduce the communication overhead by performing local processing results in increasing the energy spent on computation, and viceversa. In order to break this deadlock and make substantial improvements in the lifetime of a sensor network, there is a critical need for local computation to be available at low power, which would then allow decisions to be made locally and data to be compressed into pertinent information. Meeting this need would relieve the sensor nodes of constant-communication requirements and the associated high power-consumption costs.

 

In this project, our objective is to develop wireless sensor networks that are capable of increased computational ability at each sensor node while simultaneously increasing the longevity of operation. To do so, we will capitalize upon on the low-power and parallel computation available with programmable analog circuitry, which we will use in concert with sensor node (motes).

 

 

We have performed preliminary studies of combining analog signal processing with wireless sensor networks [1] in which we used analog circuits to perform spectral decomposition of a signal while consuming < 3μW, which is far less than the power consumed by a Crossbow TelosB mote in its lowest-power sleep mode. Compared with standard design practices for optimizing the processing-communication tradeoff that often yield only incremental improvements in energy savings, the analog signal processing based approach promises substantial improvement in the performance and longevity of the system.

 

Publications

 

o   B. Rumberg, D. Graham, V. Kulathumani, Hibernets: Energy efficient sensor networks using analog signal processing, 1(3), 321334, IEEE JETCAS, 2011

o   B. Rumberg, D. Graham, V. Kulathumani, Hibernets: Energy efficient sensor networks using analog signal processing, IPSN 2010, Stockholm, Sweden