| Embedded systems are evolving from single stand-alone entities into true networks of wirelessly connected devices, providing ubiquitous access and sensor-based information gathering. A key challenge is the management of energy in these systems, as they are battery powered for untethered operation. More and more, their energy consumption is dominated not by data processing, but by wireless communication. To ensure economic viability, this communication energy has to be reduced beyond what is achievable through digital circuit techniques, improvements in semiconductor technology, or advances in radio architectures alone.; This dissertation tackles the problem of reducing the energy consumption in wireless communications, through the use of radio power management . To this end, a link-layer framework is developed that hinges on energy-awareness, where the radio settings are adapted in the presence of variations in traffic load, channel behavior and overall network conditions. Adapting the modulation and the forward error correcting code are studied as examples. By analyzing the relative importance of electronic circuit power versus that of the power amplifier, we show that different energy management strategies are required depending on which of these two contributions is dominant.; Based on this realization, we develop a range of packet scheduling routines, dealing with both real-time and best-effort traffic, that provide energy-awareness within our framework. These routines are subdivided into two main classes, depending on the system dynamics with respect to the time variations in the wireless channel. In addition, in a lot of scenarios where data transmissions are relatively infrequent, substantial savings can be obtained by turning the radio off in the absence of traffic. The downside is that this necessitates a wakeup protocol to notify the receiver, which is also discussed in this dissertation. |
