[Paper Review] Harvesting Resource Allocation in Energy Harvesting Wireless Sensor Networks

Shenqiu Zhang, Alireza Seyedi|arXiv (Cornell University)|Jun 20, 2013

Energy Harvesting in Wireless Networks14 references3 citations

TL;DR

This paper proposes an analytical framework for optimizing energy harvesting resource allocation in wireless sensor networks with arbitrary topologies, modeling event loss probability due to energy shortages and channel impairments. It derives a closed-form expression for overall network loss and evaluates optimal, almost-fair, and uniform allocation schemes, showing the almost-fair approach achieves near-optimal performance with significantly lower energy harvesting requirements than uniform allocation.

ABSTRACT

Considering an energy harvesting sensor network, the overall probability of event loss is derived. Based on this result, a variety of harvesting resource allocation schemes (sizing the energy storages and the harvesting devices, under a total cost constraint) are provided. Their performances are verified and compared through simulations.

Motivation & Objective

• To model and analyze the overall probability of event loss in energy harvesting wireless sensor networks with arbitrary topologies.
• To formulate and solve an optimization problem for sizing energy harvesting devices and storage units under total cost constraints.
• To compare the performance of different harvesting resource allocation schemes—optimal, almost-fair, and uniform—under realistic network conditions.
• To validate the analytical loss model through extensive simulations using realistic sensor parameters and network topologies.

Proposed method

• Derives a matrix-based analytical expression for the overall network event loss probability using a left-stochastic routing matrix and energy shortage probabilities.
• Models event report generation, routing, and losses via a system of linear equations incorporating energy availability and channel impairments.
• Represents the total event rate at each node using a matrix equation involving the routing matrix R, loss probability matrix P, and channel loss q.
• Solves for the total event rate at the sink using the inverse of a matrix expression involving R, P, and q, enabling computation of overall loss probability.
• Proposes three resource allocation schemes: optimal (cost-constrained optimization), almost-fair (balanced performance and fairness), and uniform (equal allocation across nodes).
• Validates the theoretical loss model via Monte Carlo simulations over 1,482 randomly generated networks with varying node counts and parameters.

Experimental results

Research questions

• RQ1 What is the overall probability of event loss in an energy harvesting wireless sensor network with arbitrary topology, considering both energy shortage and channel impairments?
• RQ2 How can energy harvesting and storage resources be optimally allocated across sensor nodes under a total cost constraint to minimize event loss?
• RQ3 How does the performance of the proposed almost-fair resource allocation scheme compare to optimal and uniform schemes in terms of loss probability and energy harvesting requirements?
• RQ4 To what extent does the analytical loss model accurately predict real-world network behavior under stochastic energy harvesting and variable traffic?

Key findings

• The theoretical loss probability derived via matrix inversion matches simulation results closely across 1,482 randomly generated networks, validating the analytical model.
• The uniform resource allocation scheme results in a network event loss probability that is, on average, 2.2 orders of magnitude higher than the optimal scheme.
• The proposed almost-fair allocation scheme performs within 0.15 orders of magnitude of the optimal solution, demonstrating near-optimal performance with improved fairness.
• The uniform allocation requires approximately 4× higher harvesting power than the optimal or almost-fair schemes to achieve the same loss performance.
• The almost-fair scheme reaches a performance floor earlier than the uniform scheme, especially at low harvesting rates, indicating robustness under limited energy availability.
• The optimal allocation scheme achieves the lowest loss probability but requires complex optimization, while the almost-fair scheme offers a strong trade-off between performance and implementation complexity.