ASP: an energy-efficient Bluetooth polling algorithm for sensor networks

Among the various wireless technologies available today, Bluetooth stands out as the most promising for use in low-power sensor networks. This is especially true for large networks of sensors with low data rates, where Bluetooth's low power Hold, Sniff, and Park modes could be utilized during periods of low activity to reduce power consumption. One way that these modes can be exploited to reduce energy consumption is by using an intelligent polling scheme to control intra-piconet traffic. We introduce a novel polling algorithm called Adaptive Share Polling (ASP) that is designed to perform well when the network consists of sources sending short data packets at a constant rate that may fluctuate over time. In ASP, the scheduler at the master implicitly learns the share of the bandwidth that needs to be allocated to each of the slaves in order to meet the latency and/or power requirements of the application. Our simulation results show that significant power savings can be achieved in slaves whose data rates do not approach capacity of a Bluetooth piconet, and in masters when the composite data rate of the piconet does not approach capacity.

Power savings vary dramatically with the offered piconet traffic load.  The following figure shows the power savings that can be achieved in a single slave piconet by using ASP rather than a persistent polling method.  Different curves represent the choice of different polling success rates (a variable of ASP).



The price that is paid for these power savings is an increase in average packet delay.  This tradeoff can be balanced on a per node basis by tuning the polling success rate so that the application or network requirements are optimally met.  The figure below illustrates this tradeoff in a single slave piconet.  Movement along a single curve representing an application data rate is equivalent to changing the polling success rate



A complete description of ASP as well as detailed analysis and simulation results can be found in the following paper:

M. Perillo and W. Heinzelman, "ASP: An Adaptive Energy-Efficient Polling Algorithm for Bluetooth Piconets,'' Proceedings of the 36th International Conference on System Sciences (HICSS '03), January 2003.  [PDF]