Energy Harvesting Receivers: Packet Sampling and Decoding Policies

When receivers rely on stochastic energy harvesting, power outages will reduce the reliable communication rate. To model the receiver, we decompose the processing tasks in two parts: (1) sampling or Analog-to-Digital Conversion (ADC), which includes all RF front-end processing, and (2) decoding. This work considers a time-slotted system in which a source transmits coded packets through a time-invariant channel to a receiver with a finite battery and a decoder whose energy consumption can be adapted from slot to slot. Based on the relative energy costs of sampling and decoding and a stochastic model for the energy harvesting, we determine sampling and decoding policies that maximize the packet throughput. When the receiver is battery-constrained, we show that the throughput is maximized by a deferred decoding policy that samples packets at every opportunity but decodes backlogged packets using only excess energy that would otherwise be discarded. This result is shown to hold for Markov-modulated energy harvesting processes with arbitrary memory. In addition, when the average energy harvesting rate is not sufficient to decode all packets, a modified policy that defers decoding only when the backlog is below a threshold is shown to perform well when the threshold is carefully chosen.