A Unified Data-Link Energy Model for Wireless Sensor Networks

 Lizhi Charlie Zhong, 2004 Ph.D. Thesis

Advisor: Jan Rabaey

Low power consumption is the most important requirement in wireless sensor networks. Many questions arise in the design of such low-power wireless networks. How many channels are needed? Is it better to use random access or dedicated channel assignment? What is the proper data rate for the radio? When is power control effective and when should power management be used instead? How low can the power consumption be for given Quality of Service (QoS) requirements?

Analytical models provide deep insight into these key issues. For this reason, a framework of analytical models has been established in this dissertation for wireless sensor networks. Special attention is given to the energy models for the data-link layer, because its energy consumption dominates the overall energy consumption. The data-link layer models and the abstracted models of the interacting layers are integrated in this framework to expose the dependence of the overall power consumption on the design parameters of wireless sensor networks. Being integrated is what makes this framework unique. In fact, this integrated framework is very useful to the designers of wireless sensor networks: it provides venues and guidelines for energy reduction and design improvements. Furthermore, it can be a place to raise and address fundamental questions. However, the existence of the closed feedback loops makes the integration of the models very hard. Methodology of solving this closed-loop problem is given in this dissertation. The conditions imposed by the convergence on individual parts are also specified.

The integrated modeling framework presented in this dissertation uses an analytical approach, which provides designers better insight into the low-power design of the datalink layer than simulations or experiments. It is superior to other analytical approaches on that it considers the interactions between the components, enabling designers to reduce the overall power consumption even further. The validity of the models in this framework has been verified using OMNET++ network simulations. In addition, the guidelines resulting from this framework have already had a real impact on the design of an actual sensor network.