Abstract
Traditional wireless communication systems rely on stable power supplies, such as batteries or the electrical grid. However, the emergence of energy harvesting (EH) from ambient sources—solar, wind, and vibrations—presents a sustainable but highly variable alternative. This paper presents a unified mathematical framework for optimizing wireless communications powered by renewable energy. By addressing the inherent randomness of energy arrivals and fluctuating channel conditions, the framework provides a roadmap for designing efficient, battery-free communication networks through optimized power allocation and scheduling strategies.
Introduction
As the demand for ubiquitous wireless connectivity grows, the environmental and logistical costs of maintaining battery-powered devices have become significant. Energy harvesting (EH) offers a path toward self-sustaining networks, yet it introduces a fundamental challenge: instability. Unlike conventional systems, EH-powered transmitters must operate under an "energy causality" constraint, meaning they cannot consume energy before it is harvested. Furthermore, these systems must manage a multi-time-scale mismatch, where energy harvesting rates fluctuate over seconds or minutes, while wireless channel conditions vary on a millisecond scale. This paper develops a general utility optimization framework to jointly manage these dynamics.
Problem Formulation and Methodology
The core of the research is the development of a general optimization problem, designated as (P1), which seeks to maximize the sum of utility functions over a finite time horizon. The framework is designed to be flexible, allowing the "utility" to represent various performance metrics, such as Shannon capacity (throughput), non-outage probability, or end-to-end distortion.
To address the uncertainty of future states, the authors categorize the transmitter's knowledge into four distinct information scenarios:
- Non-causal: The transmitter has perfect knowledge of past, present, and future energy and channel states.
- Causal: The transmitter knows the past and present but relies on statistical distributions to predict the future.
- Hybrid: The transmitter has causal knowledge of the channel (CSI) but non-causal knowledge of energy arrivals (ESI).
- No CSIT: The transmitter has no channel state information at all.
The study employs convex optimization and Karush-Kuhn-Tucker (KKT) conditions for non-causal scenarios, while Dynamic Programming (DP) and Markov Decision Processes (MDP) are used to solve causal scenarios involving stochastic variables.
Key Insights and Transmission Strategies
The framework reveals several optimal transmission strategies tailored to specific information and utility constraints:
The Staircase Water-Filling Effect
For throughput maximization with perfect future knowledge (non-causal), the optimal power allocation follows a "staircase water-filling" structure. In this model, the power level is non-decreasing over time. As the transmitter harvests more energy and approaches its operational deadline, it can afford to be more aggressive with its power consumption, effectively "filling" the available channel capacity as energy becomes available.
Save-then-Transmit
In scenarios where the transmitter lacks channel knowledge (No CSIT) and faces a high risk of outage, a "save-then-transmit" strategy is proven optimal. Rather than transmitting immediately with low energy, the transmitter harvests and stores energy until it reaches a critical threshold ($P_c$). This buffer ensures that once transmission begins, the link can be maintained with sufficient reliability to minimize dropped packets.
Energy Cooperation and WET
In multiuser and relay environments, the framework introduces the concept of Wireless Energy Transfer (WET). This allows an energy-rich node to transmit electricity over the air to an energy-poor node. This "energy cooperation" balances the network's energy state, significantly boosting total system throughput compared to systems where nodes operate in isolation.
Results and Performance Bounds
The theoretical analysis establishes that non-causal knowledge provides a performance upper bound for all other scenarios. In the simplest Additive White Gaussian Noise (AWGN) case with a constant channel, the optimal power is a strictly non-decreasing staircase function. For fading channels where the outage probability function is non-convex, the "save-then-transmit" structure is shown to be the most effective method for maintaining link integrity. Additionally, in three-node relay systems, the authors demonstrate that if the destination can tolerate delays, the power allocations at the source and relay can be decoupled and optimized independently, simplifying the design of complex EH networks.
Limitations and Future Directions
While the framework provides a robust foundation for EH-based communications, several areas remain for future exploration:
- Receiver Energy Costs: Current models focus primarily on the transmitter, but the energy required for decoding at the receiver is significant and must be integrated into future optimization problems.
- Cross-Layer Design: The assumption of "backlogged" data (always available) should be replaced with models that account for the random arrival of data packets alongside random energy arrivals.
- Hardware Realities: Future iterations must account for finite battery capacity and energy leakage, moving beyond the "ideal storage" models used in current theoretical frameworks.
- RF Energy Harvesting: The use of dedicated radio frequency beams for power (WET) requires more sophisticated beamforming and interference management to be viable in dense network deployments.
Summary updated