Confidence-based data management for personal area sensor networks

The military is working on embedding sensors in a "smart uniform" that will monitor key biological parameters to determine the physiological status of a soldier. The soldier's status can only be determined accurately by combining the readings from several sensors using sophisticated physiological models. Unfortunately, the physical environment and the low-bandwidth, push-based personal-area network (PAN) introduce uncertainty in the inputs to the models. Thus the model must produce a confidence level as well as a physiological status value. This paper explores how confidence levels can be used to influence data management decisions. In particular, we look at power-efficient ways to keep the confidence above a given threshold. We also contrast push-based broadcast schedules with other schedules that are made possible by two-way communication.

• Type: Conference
• Year: 2004
• Where: DMSN
• Authors: Nesime Tatbul, Mark Buller, Reed Hoyt, Steve Mullen, Stanley B. Zdonik
• Keywords: Confidence Levels, DMSN 2004, Physiological Status, Sensor Networks, Sophisticated Physiological Models