Quasi-static Centralized Rate Allocation for Sensor Networks

Rate control for congestion mitigation and avoidance has received significant attention in the sensor networks literature. Existing rate control schemes dynamically assign rates in a distributed manner. In this paper, we take a step back and ask: is a near-optimal quasi-static centralized rate allocation even feasible for wireless sensor networks? Intuition would suggest otherwise, since wireless conditions vary dynamically, and optimal centralized rate allocation is known to be computationally intractable. Surprisingly, however, we find that, quasi-static centralized rate allocation performs well at time-scales of tens of minutes on a 40-node testbed. Our approach relies on a relatively simple, lightweight rate allocation heuristic that uses topology and loss rate information, and adapts at relatively long time-scales to channel variability. Extensive experiments on a 40-node wireless testbed show that sensor nodes achieve a goodput very close to their allocated rate, even in harsh wireless conditions. Furthermore, this achieved goodput is nearly 50% higher than that achieved by IFRC, a recently-proposed distributed rate control scheme, and within 13% of an empirically-determined optimal rate. We also evaluate extensions to our heuristic to support weighted fairness and networks with multiple base stations.