Achieving Near-Optimal Fairness in Energy Harvesting Wireless Sensor Networks

This paper tackles a single-hop wireless sensor network (WSN) where a fusion center (FC) receives data from M energy harvesting (EH) wireless sensors. Each sensor stores its harvested energy losslessly in an infinite capacity battery. In each time slot, K sensors are selected for data transmission over K mutual orthogonal channels. FC knows neither battery states of sensors, nor statistics of EH processes at sensors. FC just knows outcomes of previous transmission attempts. A sensor transmits data if it is scheduled and has enough energy in its battery. FC aims to maximize fairness among EH sensor nodes over the time horizon under the proportionate progress (P-fairness) by using Jain's Fairness Index. Uniforming Random Ordered Policy (UROP) and Myopic Policy (MP) are suggested as solutions to the fairness problem. It is shown both analytically and numerically that under infinite capacity battery assumption, UROP achieves asymptotically optimal performance in fairness under quite general EH process over infinite time horizon. Even with finite capacity batteries, UROP can achieve nearly optimality in fairness over finite time horizons whereas the MP is generally suboptimal.