Efficient routing from multiple sources to multiple sinks in wireless sensor networks

Initial deployments of wireless sensor networks (WSNs) were based on a many-to-one communication paradigm, where a single sink collects data from a number of data sources. Recently, however, scenarios with multiple sinks are increasingly being proposed, e.g., to deal with actuator nodes or to support high-level programming abstractions. The resulting many-to-many communication makes the existing solutions for single-sink scenarios inefficient. In this paper, we propose a scheme for routing data efficiently from multiple sources to multiple sinks. We first study the problem from a theoretical standpoint, by mapping it to the multi-commodity network design problem. This allows us to derive an optimal solution that, albeit based on global knowledge and therefore impractical, provides us with a theoretical lower bound to evaluate decentralized solutions against. Then, we propose our own decentralized scheme, based on a periodic adaptation of the message routes aimed at m minimizing the number of network links exploited. The resulting protocol is simple and easily implementable on WSN devices. The evaluation of our implementation shows that our protocol generates 50% less overhead than the base scheme without adaptation, a result close to the theoretical optimum we derived.