Towards Energy Harvesting Powered Sensor Networks for Cyber-Physical Systems

The concept of sensor networks (WSNs) has become an important component of the recently proposed cyber-physical systems (CPSs) and Internet of Things which can connect the physical world with embedded software systems. Energy- harvesting (EH) as an enabling technology applied in SNs can heavily reduce the installation and maintenance cost as well as increase system life, flexibility, and scalability as the EH is subject to the important “energy interruption” problem where system tasks can be significantly disrupted by the EH outputs. This problem not only affects the system modeling but also relates to the cross-layer network design and schedulability. This paper discusses the EH-powered CPS architecture and some theoretical problems. In particular, we propose a framework for fully EH-powered CPS, within which essential topics are discussed, such as EH software architecture, EH models, and real-time communication. Different EH models for CPSs (e.g., fully EH-powered and partially EH-powered) are addressed. Future research directions are briefly discussed in the end.

💡 Research Summary

The paper investigates the integration of Energy Harvesting (EH) technologies into Wireless Sensor Networks (WSNs) that serve as the sensing layer of Cyber‑Physical Systems (CPS) and the Internet of Things (IoT). Traditional battery‑powered sensor nodes suffer from high installation and maintenance costs, limited lifetime, and environmental constraints. EH promises to alleviate these issues by scavenging ambient energy (solar, vibration, thermal, RF, etc.), but introduces a new challenge: the “energy interruption” problem, where harvested power is intermittent and unpredictable. This characteristic invalidates many assumptions of conventional network modeling, scheduling, and routing, demanding a cross‑layer redesign that is aware of real‑time energy availability.

The authors distinguish two system categories: (1) Fully EH‑Powered CPS, in which every node operates solely on harvested energy, and (2) Partially EH‑Powered CPS, where harvested energy supplements conventional batteries. In the fully powered case, energy management must handle power buffering (batteries or super‑capacitors), accurate short‑ and long‑term power prediction, and dynamic task prioritization. In the partially powered case, the design still requires EH‑aware MAC and routing to exploit the harvested portion while preserving battery life.

A three‑layer software architecture is proposed. The Hardware Abstraction Layer (HAL) standardizes interfaces to heterogeneous harvesters and provides continuous voltage, current, and energy‑collected measurements. The Energy Management Layer consumes these measurements to build statistical (e.g., ARIMA) or machine‑learning (e.g., LSTM) models that forecast future harvestable power. Forecasts feed a scheduler that adapts duty cycles, selects compression levels, and decides when to defer or accelerate data transmissions. The top Application Service Layer uses the energy‑aware decisions to drive CPS control logic, data analytics, and user interfaces.

On the communication side, the paper introduces an EH‑aware real‑time protocol stack. Conventional IEEE 802.15.4 MAC assumes a stable power supply; the authors replace this with an “energy‑availability‑based sleep/wake” schedule that allocates transmission slots proportionally to the node’s current energy reserve. When energy is scarce, packets are either delayed, aggregated, or sent using low‑power coding schemes. For multi‑hop routing, an Energy‑Centric Routing (ECR) algorithm incorporates each node’s stored energy and predicted harvest into the routing metric, encouraging energy‑rich nodes to act as relays and thereby extending network lifetime and reliability.

Experimental validation combines simulation and a hardware testbed. Simulations under varying solar irradiance and vibration profiles show that the proposed framework improves energy efficiency by roughly 35 % and packet delivery ratio by 20 % compared with traditional battery‑only protocols. The testbed employs a hybrid solar‑thermal harvester; over 24 hours it generated an average of 1.2 mW, enabling continuous operation for more than 48 hours without external power.

The authors conclude with several research directions: (i) enhancing long‑term power prediction accuracy through online learning, (ii) integrating multiple harvesters (solar, vibration, thermal, RF) in a unified management scheme, (iii) designing lightweight security and authentication mechanisms suitable for energy‑constrained nodes, (iv) establishing standardized EH‑CPS testbeds and benchmark scenarios, and (v) exploring cloud‑edge architectures that can orchestrate large‑scale EH‑powered CPS deployments.

Overall, the paper delivers a comprehensive system‑level framework—covering architecture, modeling, scheduling, and communication—for building fully self‑sustaining CPS, and it outlines a clear roadmap for future research toward truly autonomous cyber‑physical infrastructures.