Design of a real-time adaptive power optimal system

Wireless sensor systems are currently being deployed in a wide variety of lightweight mobile applications such as distributed object tracking and wearable medical data collection. For such applications to enter the consumer mainstream, it is necessary for them to operate far more power efficiently than they do currently. Using a simple illustrative example, this paper explores some general design rules for techniques which can reduce power consumption by one to two orders of magnitude through a combination of real-time sensor selection based on system state and in-situ feature extraction before storage and/or transmission (if necessary). In the presented example, it is demonstrated that the use of both low and high accuracy sensors for a single parameter (acceleration) can allow system state identification to be far more power-efficient (in this case, by 94%) than with a single sensor alone. Use of sensors with short wakeup times will further improve this result.