Finite-horizon online throughput maximization for an energy harvesting transmitter

An online finite horizon throughput maximization problem is considered. Specifically, optimal power allocation of an energy harvesting rechargeable node with two different accessible output power levels is studied. Communication takes place under a static channel and rate levels are assumed to be a concave function of power, implying the delay-energy efficiency tradeoff. Taking battery state, remaining number of slots and achievable rate levels into account optimal policy is obtained through dynamic programming. In addition, several policies that are suboptimal yet good are proposed. Based on an empirically motivated harvesting model, a policy that we call Expected Threshold Policy is shown to achieve near-optimal performance.