The design of power-aware embedded systems.

摘要

This dissertation explores various research issues associated with power aware embedded systems in energy-limited environment. The ever-growing gap-called “battery gap”—between slowly improving battery capacity and rapidly increasing energy requirements of embedded systems makes it crucial to utilize energy resources efficiently, thereby increasing the system's lifetime (usefulness) and narrowing the battery gap. In this dissertation we investigate system level power aware approaches that will maximize the energy efficiency in energy-limited environment by exploiting the interaction among different components (resources).; In this dissertation we take two different approaches, the network level approach and the node level approach, in achieving energy efficiency. At the network level, we maximize the energy utilization by studying the computation/communication trade-off in wireless embedded systems. By formulating the computation/communication trade-off strategy as a linear fractional programming problem, we derive optimal strategy to allocate computation/communication resource to different tasks. At the node level, we introduce novel battery state aware approaches which maximize battery utilization by dynamically adapting the system performance in accordance to changing battery state. Our results, obtained through experiments using real-life batteries, demonstrate that such a battery-state aware approach to energy management leads to significant improvements in total service produced by the system.; We also discuss design of a novel power aware wireless sensor node platform called “iBadge”. With the use of power tracking and management unit, iBadge provides a versatile platform in which various power aware algorithms can be tested and implemented. A novel sensor network simulation environment called “SensorSim” is discussed. SensorSim builds upon a conventional network simulation tool by providing abilities to create realistic sensing scenarios. SensorSim also provides users with abilities to study power consumption in sensor networks under various protocols and power management schemes.