Skip to Main Content
This paper presents an optimization framework for a wireless sensor network in which each sensor plays a dual role of sensing the environment and relaying the sensor information. The design of such a network involves two distinct aspects. First, as the observations of the underlying environment are often correlated, distributive source coding methods have the potential to greatly improve the efficiency of the sensor operation. Thus, information theoretical source coding methods are useful in the application layer. Second, as each sensor must send information individually to a central processor, routing and power allocation in the network and physical layers are also important issues. The main focus of this paper is an optimization framework that jointly solves the source coding, routing and power allocation problems in such a network. The main insight is the following: the joint optimization problem for a sensor network, when solved in the dual domain, provides a natural separation between the application layer, the network layer and the physical layer. The interface between the layers is precisely the dual optimization variables. The crucial observation that makes this possible is that the underlying source coding problem in the application layer and the channel coding problem in the physical layer can always be made convex via time-division or frequency-division multiplexing. Convexification in time or frequency enables dual algorithms to reach the global optimum of the overall network optimization problem efficiently.