Asymptotically Optimal Uncoordinated Power Control Policies for Energy Harvesting Multiple Access Channels With Decoding Costs

The objective of this paper is to design a power control policy that maximizes the long-term time-averaged sum throughput of a Gaussian multiple access channel (MAC), where the transmitters as well as the access point (AP) are energy harvesting nodes (EHNs). In addition, the policy is required to facilitate uncoordinated operation of the network. That is, in each slot, the transmitting nodes and the AP need to independently take their actions, e.g., the amount of energy to be used for transmission or whether to turn on and receive the data. First, in order to benchmark the performance of any policy, we derive an upper bound on the throughput achievable, by analyzing a centralized genie-aided system where the nodes have infinite capacity batteries and can freely share the available energy among themselves. In addition, the genie-aided system has non-causal knowledge of the energy arrivals at all the nodes. Next, we show that, surprisingly, a simple time sharing based online policy which requires no coordination among the transmitters and uses time-dilation at the receiver achieves the upper bound asymptotically in the battery size. We also present a policy that requires an occasional one-bit feedback from the AP about its battery state, and show that it requires a smaller sized battery at the receiver compared to a policy which operates without any feedback from the AP, to achieve the same performance. We use Monte Carlo simulations to validate our theoretical results and illustrate the performance of the proposed policies.