Optimal energy allocation for wireless communications powered by energy harvesters

We consider the use of energy harvesters, in place of conventional batteries with fixed energy storage, for point-to-point wireless communications. In addition to the challenge of transmitting in a channel with time selective fading, energy harvesters provide a perpetual but unreliable energy source. In this paper, we consider the problem of energy allocation over a finite horizon, taking into account a time varying channel and energy source, so as to maximize the throughput. Two types of side information are assumed to be available: causal side information of the immediate past channel condition and harvested energy, and full side information. We obtain structural results for the optimal energy allocation, via the use of dynamic programming and convex optimization techniques. In particular, if unlimited energy can be stored in the battery with harvested energy, we prove the optimality of a water-filling energy allocation solution with multiple, non-decreasing water levels.