Abstract
In this paper, we derive optimal transmission policies for energy harvesting sensors to maximize the utility obtained over a finite horizon. First, we consider a single energy harvesting sensor, with discrete energy arrival process, and a discrete energy consumption policy. Under this model, we show that the optimal finite horizon policy is a threshold policy, and explicitly characterize the thresholds, and the thresholds can be precomputed using a recursion. Next, we address the case of multiple sensors, with only one of them allowed to transmit at any given time to avoid interference, and derive an explicit optimal policy for this scenario as well.
Introduction
Wireless sensor nodes powered by energy harvesting (such as solar or vibration) face a fundamental "save or spend" dilemma. Because energy arrivals are stochastic and battery capacity is finite, a node must decide whether to transmit immediately to exploit a favorable channel or conserve energy for future high-priority data or better channel conditions. While long-term average optimization (infinite-horizon) is well-studied, many real-world sensing tasks operate within a specific "finite-horizon" window where data utility is time-sensitive. Solving for optimal policies in these scenarios with random energy bursts and time-varying channels is computationally intensive.
Methodology
The authors employ Dynamic Programming (DP) to derive explicit, computable policies. A key innovation is modeling both energy arrivals and consumption as discrete quanta, which transforms the continuous optimization problem into a tractable set of recursive equations.
Single Node Case
The study first addresses a sensor with Bernoulli energy arrivals and binary consumption, subsequently generalizing this to any discrete energy model. The optimal strategy is proven to be a threshold policy. At each time step, the sensor compares the immediate reward of transmission against recursive thresholds ($\gamma$) representing the "future value" of stored energy. These thresholds can be precomputed offline.
Multiple Node Case
The model is extended to multiple nodes sharing a single channel where only one can transmit at a time. To address the "curse of dimensionality" in joint optimization (where the state space grows exponentially with the number of nodes and battery size), the authors propose a decoupled sub-optimal policy. In this heuristic, each node independently calculates its own threshold, and the node with the best channel state is permitted to transmit.
Results
Numerical simulations were conducted to validate the theoretical findings. For the single-node case, the performance of general versus binary energy arrival models was compared, demonstrating how the total payoff scales with the number of available time slots.
In the multi-node scenario, the performance of the complex joint optimal policy was compared against the proposed decoupled sub-optimal policy. The results indicate that the performance gap is negligible, supporting the use of simpler policies in resource-constrained hardware.
Discussion and Conclusion
The research highlights that discretization is the key to making finite-horizon optimization tractable. The derived thresholds act as a "price" for energy; if the current channel utility exceeds this price, transmission is optimal. Crucially, the near-optimality of the decoupled policy suggests that complex central controllers are not necessary for multi-sensor systems. Instead, decentralized nodes acting on local state information can achieve high efficiency, significantly reducing the computational burden on low-power sensor designs. Future work may address non-linear battery characteristics and imperfect channel state information.
Summary updated