Introduction: The Challenge of Stochastic Energy Harvesting
In the era of "green" communications and the Internet of Things (IoT), wireless nodes are increasingly deployed in remote environments where they must rely on energy harvesting (EH) from sources like solar power. While sustainable, solar energy is inherently unpredictable, fluctuating with weather patterns and diurnal cycles.
Optimizing performance in such networks is particularly complex for Two-Way Relay (TWR) networks. In a TWR setup, two source nodes exchange information through a central relay. Because the communication is bidirectional, the failure of a single node—due to energy depletion—renders the entire exchange unsuccessful. This paper addresses the challenge of coordinating transmission policies across multiple nodes under random energy arrivals, finite battery capacities, and fading wireless channels.
System Model and Problem Formulation
The researchers consider a TWR network consisting of two source nodes ($S_1$ and $S_2$) and one relay ($R$). The communication occurs in two phases: the Multiple Access (MA) phase, where sources transmit to the relay, and the Broadcast (BC) phase, where the relay forwards the combined information back to the sources.
The study explores two primary scenarios for the relay:
- Fixed Power (FP) Mode: The relay has a stable, external power supply.
- Energy Harvesting (EH) Mode: The relay, like the source nodes, must harvest solar energy to function.
Methodology: An MDP Framework
To find the optimal transmission strategy, the authors formulate the problem as an infinite-horizon Markov Decision Process (MDP). The goal is to maximize the network's long-term average reward, defined as the complement of the outage probability (i.e., the probability of successful communication).
The system state is defined by five key variables: the current solar irradiance state (modeled using a Gaussian mixture hidden Markov model based on real-world data), the fading conditions of the two wireless channels, and the battery levels of the source nodes.
The optimal policy is derived by solving the Bellman equation using Value Iteration. This process maps every possible combination of system states to the transmission power levels that minimize outages over time.
Key Insights and Optimal Policy Structure
The analysis reveals several elegant properties of the optimal transmission policy:
- Threshold-Based Decisions: The optimal policy follows an "on-off" threshold structure. In FP mode, this is a 2D threshold based on the battery levels of the two sources. In EH mode, it expands into a 3D threshold that includes the relay's battery state.
- The "Zero-Action" Property: A critical finding is that if one source node’s battery is depleted, the optimal action for the other node is to remain silent. Since both sides must succeed for a valid two-way exchange, transmitting alone would simply waste harvested energy that could be saved for a future slot when both nodes are ready.
- Power Convergence: Under high Signal-to-Noise Ratio (SNR) conditions, nodes do not simply use maximum power. Instead, they converge to a "basic power unit," balancing current transmission success against the need to maintain battery reserves for future periods of low solar availability.
- Deficiency Probability: The authors identify that the ultimate performance limit is dictated by the "deficiency probability"—the likelihood that the network's batteries are too low to support any transmission, regardless of channel conditions.
Results and Performance Evaluation
The proposed MDP-based policy was validated against Monte-Carlo simulations and compared to several heuristic strategies, including greedy "on-off" policies and "power halving" (using half the available battery).
The results demonstrate that the optimal policy significantly outperforms heuristics by intelligently managing energy reserves. Furthermore, while the EH relay mode introduces more uncertainty and a higher outage floor than the FP mode, the authors show that increasing the relay's solar panel size can effectively mitigate this performance gap.
Conclusion
By moving beyond idealized "non-causal" energy models and utilizing real-world solar data, this research provides a practical framework for managing energy-constrained TWR networks. The identified threshold structures and the "zero-action" property offer clear, implementable rules for maximizing the reliability of green wireless communications in unpredictable environments.
Reference: Wei Li, et al., "Transmission Policy of Two-Way Relay Networks With Multiple Stochastic Energy Harvesting Nodes," in IEEE Access, vol. 7, 2019. DOI: 10.1109/ACCESS.2019.2922222.
Summary updated