Abstract
This paper addresses the challenge of optimal energy and bandwidth allocation in fading channels for multiple energy-harvesting (EH) transmitters. We formulate a convex optimization problem to maximize the weighted sum-rate over $K$ time slots, considering finite battery capacity and peak power constraints. To overcome the high computational complexity of generic solvers, we propose an iterative algorithm that decomposes the problem into energy and bandwidth allocation subproblems. For the special case of multiple point-to-point channels, we derive efficient algorithms with $O(MK^2)$ complexity. Additionally, an adaptive water-filling heuristic is presented for scenarios with only causal information.
1. Introduction
In the evolving landscape of the Internet of Things (IoT), energy harvesting (EH) from ambient sources like solar or RF signals is a key enabler for sustainable communication. However, EH introduces unique scheduling complexities:
- Time-Coupling: Harvested energy can be stored for future use, but finite battery capacities necessitate careful management to avoid energy overflow.
- Space-Coupling: Multiple transmitters sharing a frequency band must coordinate bandwidth allocation to maximize network throughput without mutual interference.
While prior work often treats these dimensions in isolation, this research tackles the joint allocation of energy and bandwidth across multiple time slots to maximize the network's weighted sum-rate.
2. System Model and Problem Formulation
The system consists of $N$ transmitters, each equipped with an energy harvester and a finite battery. Each transmitter $n$ communicates with a set of receivers $M_n$ via orthogonal channels.
The objective is to determine the transmission power $p_m^k$ and bandwidth $a_m^k$ for each receiver $m$ in each time slot $k$ to maximize the total throughput. The problem is constrained by energy causality, battery capacity limits, and maximum instantaneous power constraints.
3. Proposed Methodology
The joint optimization is a convex problem, but the $O(MK)$ constraints make standard solvers impractical for real-time use. The authors propose an iterative decomposition approach:
3.1 Iterative Decomposition
The algorithm alternates between solving two subproblems:
- Energy Allocation (EP): With bandwidth fixed, the transmitter optimizes power distribution over time using a "Discounted Dynamic Water-Filling" approach.
- Bandwidth Allocation (BP): With power fixed, the central controller determines the optimal frequency shares for each user.
3.2 Key Insights
A critical finding is the inefficiency of TDMA in EH systems. In traditional power-constrained systems, Time Division Multiple Access (TDMA) is often optimal. However, in EH systems, preventing battery overflow requires simultaneous bandwidth sharing to "drain" energy when harvesting rates are high. The authors prove that reaching "pairwise optimality" between these subproblems guarantees a global optimum.
4. Performance Evaluation
The algorithms were evaluated using Monte Carlo simulations with Rayleigh fading channels and truncated Gaussian energy harvesting models.
- Efficiency: The iterative algorithm typically converges to the global optimum within 4 to 7 iterations.
- Complexity: By exploiting KKT conditions, the complexity is reduced to $O(NK^2)$, significantly lower than generic convex solvers.
- Causal Heuristic: For real-time scenarios where future channel states are unknown, the proposed "Adaptive Water-Filling" heuristic significantly outperforms standard "Greedy" or "Equal Bandwidth" policies.
5. Conclusion
This work provides a robust framework for joint resource allocation in EH-powered multiuser systems. By decomposing the problem, the proposed algorithms achieve optimal performance with practical computational overhead. Future work will extend these concepts to multiuser broadcast channels and non-orthogonal access schemes where interference management becomes a primary factor.
Summary updated