Maximizing transported data before partition in a wireless sensor network

We consider a structured sensor network in which sensors and relays are arranged in a regular grid. Sources S₁, S₂, ..., S_K are located sequentially at the top row, and destinations D₁, D₂, ..., D_K are arranged in an arbitrary order at the bottom row; data from S_i is required at D_i. Such networks serve as models for planned sensor deployments (for example, sensors deployed in a precision agriculture project). Nodes have limited energy, and our objective is to maximize the aggregate bits transported from source(s) to destination(s) before network partition. In contrast to formulations that seek to optimize either the network lifetime or the throughput, our criterion considers both aspects; the criterion is natural and novel. We show that our problem can be converted to a Linear Program. Next, we move on to consider relay nodes that not only forward data but are also capable of XOR-ing received packets. We study how this “coding capability” can be leveraged to improve the performance objective. We provide an algorithm that chooses the optimal number of coding points. Finally, we evaluate the additional benefit that XOR-ing can yield. For this evaluation, we formulate an optimization problem using “functional flow conservation.” Two XOR-ed packets get “fused” to one, because of which, at the coding points, the usual flow conservation notion does not apply; hence the need for functional flow conservation. We propose functional flow equations for directed acyclic graph networks; this is novel, as the current literature considers only tree graphs. Our results indicate that the gain in bits transported can be as much as 83%.