Trading Latency For Load Balancing In Many-To-One Wireless Networks

Wireless networks that combine a many-to-one traffic flow with multi-hop communication suffer from a funnelling effect that inevitably overburdens nodes closer to the base station. In sensor networks this is known as the energy hole problem. A routing protocol that constructs a shortest path tree exacerbates this problem by leaving some of the most critical nodes with more descendants than others and hence more relaying work to do.In this paper we propose a novel, fully distributed tree construction algorithm for building a load-balanced tree. Our algorithm, Degree Constrained Routing (DECOR), trades off latency for load balancing. We simulated the algorithm using a sensor network as an example. The results show that a small increase in latency, of less than 10%, can be sacrificed for a larger increase in balance of up to 80%. The lifetime of a sensor network using DECOR can be extended by up to 150% compared to the next best algorithm.