Abstract
Energy harvesting (EH) communication systems are pivotal for sustainable wireless networks but are constrained by the unpredictable arrival of energy and data. This paper addresses the problem of minimizing data completion time under causal information constraints. By modeling energy and data arrivals as continuous functions, we prove that the optimal competitive ratio for any online algorithm is exactly 2. We propose two online strategies that achieve this theoretical limit and demonstrate via numerical simulations that continuous modeling provides superior efficiency compared to traditional discretized models.
Introduction
Energy Harvesting (EH) systems, such as solar-powered sensors, are essential for "green" and long-lasting wireless networks. A primary objective in these systems is to minimize the "completion time"—the time required to transmit a specific batch of data. However, transmitters typically only have causal information, meaning they know past and present arrivals but not future ones. Most existing literature focuses on offline scenarios or discrete models. This work bridges the gap by analyzing online algorithms in a continuous arrival framework.
Methods and Algorithms
The authors model both energy and data arrivals as general continuous functions. Two online algorithms are proposed:
- Algorithm 1: A sophisticated strategy where the transmitter waits for a minimum threshold of energy and data before starting. It dynamically adjusts power by balancing buffered energy against remaining data.
- Algorithm 2: A "wait-then-go" strategy where the transmitter waits until all data has arrived and sufficient energy is available to complete the transmission in one burst.
The performance of these algorithms is evaluated using the competitive ratio—the ratio of online completion time to the offline optimal completion time. The study establishes fundamental lower bounds using game-theoretic concepts.
Key Insights and Numerical Results
The central finding is the "Power of 2": the optimal competitive ratio is exactly 2. No matter how clever an online algorithm is, if it has no knowledge of future arrival distributions, it cannot guarantee a completion time better than twice the offline optimal.
Numerical simulations validate these theoretical bounds. Using a log-type capacity function for an AWGN channel, the researchers compared continuous arrival models against discretized versions.
The simulations revealed that discretization imposes a penalty on system efficiency, increasing completion time. In practical scenarios, the proposed online algorithm achieved a ratio of 1.24, significantly better than the worst-case bound of 2. Furthermore, these algorithms are distribution-agnostic, making them robust for unpredictable real-world environments.
Conclusion
This research establishes that a well-designed online strategy for continuous EH systems is at most 2-competitive. While Algorithm 1 generally outperforms Algorithm 2 by starting transmission earlier, both meet the theoretical optimality bound. Future work may extend these findings to multi-user or relay-based networks and consider hardware-specific constraints like battery leakage.
Summary updated