Optimizing Flows in Changing Tree-based Sensor Networks

Military sensor networks often operate in resource challenged environments. This poses the problem of how to allocate resources to sensors flow to accomplish a mission. In this paper we consider a set of sensors that communicate observations up a tree to a fusion center. The value of the mission is modeled by a separable increasing concave functions and we develop a low complexity one step algorithm that allocates link capacities to each sensor so as to maximize this function. By limiting ourselves to a tree topology, we derive several important benefits, including the ability to quickly adapt to changes in utility functions or topology, and in a straightforward way to run our algorithm in a parallel, distributed manner over the network with little communication overhead and no centralized planning.