Non-Asymptotic Achievable Rates for Gaussian Energy-Harvesting Channels: Save-and-Transmit and Best-Effort

Jing Yang, Silas L. Fong, Jing Yang 0002, Aylin YenerIEEE Trans. Information Theory 2019 · DOI

TL;DR

This paper investigates the achievable communication rates for energy-harvesting (EH) wireless channels in a finite blocklength regime. The authors introduce new non-asymptotic achievable rates for "save-and-transmit" and "best-effort" communication strategies, showing they approach the theoretical capacity with improved performance, especially in high signal-to-noise ratio (SNR) scenarios, compared to prior work.

What Problem Does This Paper Address?

Wireless devices powered by energy harvesting (EH) face a fundamental challenge: the intermittent and unpredictable nature of energy availability. This makes traditional communication strategies, which assume a constant power supply, inefficient or impossible. The paper focuses on the Gaussian Energy-Harvesting (EH) channel, a standard model for such systems, and specifically addresses how to maximize data transmission rates when communication must occur within a finite block of time (finite blocklength regime). Prior work established the theoretical capacity of such channels but lacked precise, non-asymptotic achievable rates for practical schemes, particularly for the "best-effort" strategy. This paper aims to fill this gap by deriving new, tighter bounds on achievable rates for both "save-and-transmit" and "best-effort" strategies, considering both independent and block-wise energy arrivals.

Methods

The core of this paper lies in deriving new non-asymptotic achievable rates for two EH communication strategies:

  • Save-and-Transmit: This strategy involves an initial "saving phase" where harvested energy is stored in a battery, followed by a "transmission phase." A key innovation here is analyzing this scheme with a transmit power strictly less than the battery recharge rate. This "backoff" strategy, while seemingly counterintuitive, is shown to improve performance in certain regimes.
  • Best-Effort: This is a simplified version of save-and-transmit where there is no initial saving phase. The transmitter uses available energy as it arrives.

The authors employ techniques from information theory, particularly focusing on finite blocklength analysis. They utilize probability bounds derived from analyzing Markov processes and their escape probabilities. The proofs involve intricate mathematical derivations, including the use of Berry-Esséen theorems for bounding error probabilities and careful analysis of energy constraints. The paper also extends these results to a "block energy arrival model," where energy arrives in blocks rather than individual time slots, making it more applicable to scenarios with slower energy harvesting dynamics.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and does not involve empirical datasets.

Key Insights and Ideas

  • Transmit power backoff can be beneficial: The most striking insight is that allowing the transmit power to be less than the average energy recharge rate (P) in the save-and-transmit scheme can lead to higher achievable rates, especially in high SNR regimes. This is contrary to intuition that one should always transmit at maximum possible power.
  • Improved non-asymptotic rates: The paper provides new, tighter non-asymptotic achievable rates for both save-and-transmit and best-effort schemes. These rates offer a more precise understanding of performance limits in finite blocklength scenarios.
  • Optimal scaling for best-effort in block arrivals: For the block energy arrival model, the best-effort scheme achieves an optimal second-order scaling of $O(\sqrt{\max\{\log n, L\}/n})$, which is a significant improvement over previous results.
  • Unified framework: The save-and-transmit scheme with an arbitrary saving phase length (including zero) provides a unified framework that encompasses the best-effort scheme as a special case.
  • Block energy arrivals are important: The generalization to block energy arrivals highlights the practical relevance of the work for scenarios where energy harvesting is not instantaneous but occurs in discrete, longer intervals.

Key Results

The paper derives new non-asymptotic achievable rates for the save-and-transmit and best-effort schemes in the finite blocklength regime for the AWGN EH channel.

  • i.i.d. Energy Arrival (L=1):
    • Save-and-transmit achieves a second-order scaling of $O(1/\sqrt{n})$.
  • Best-effort achieves a second-order scaling of $O(\log n / n)$.
  • The proposed save-and-transmit scheme with transmit power backoff outperforms the state-of-the-art in high SNR regimes.
  • Block Energy Arrival (L = o(n)):
    • Save-and-transmit achieves a second-order scaling of $O(\sqrt{L/n})$.
  • Best-effort achieves a second-order scaling of $O(\sqrt{\max\{\log n, L\}/n})$.
  • In the case where $L = \omega(\log n)$ and $L = o(n)$, the best-effort scheme achieves an optimal second-order scaling of $O(\sqrt{L/n})$.

Limitations and Future Work

  • Finite Battery: The current analysis assumes an infinite-sized battery. The authors note that the finite battery case is an active area of research and a natural extension of this work.
  • Best-Effort Optimality: While the paper provides improved rates for best-effort, it acknowledges that for the i.i.d. energy arrival case (L=1), the achieved second-order scaling of $O(\log n / n)$ is not optimal. The optimal scaling is $O(1/\sqrt{n})$. The authors suggest future work could involve proving sharper probability bounds to potentially improve this.
  • General EH Models: The paper focuses on the AWGN EH channel. Extending these results to other channel models or more complex EH processes could be a future direction.
  • Practical Implementation: While the theoretical results are significant, translating these complex coding schemes into practical, low-complexity implementations for real-world EH devices remains a challenge.

Summary updated

BibTeX
@article{yang2019non,
  author = {Jing Yang and Silas L. Fong and Jing Yang and Aylin Yener},
  title = {{Non-Asymptotic Achievable Rates for Gaussian Energy-Harvesting Channels: Save-and-Transmit and Best-Effort}},
  journal = {IEEE Trans. Information Theory},
  year = {2019},
  doi = {10.1109/tit.2019.2927006},
}
Figure