Author ORCID Identifier

https://orcid.org/0009-0002-9393-2028

Date Available

4-28-2025

Year of Publication

2025

Document Type

Doctoral Dissertation

Degree Name

Doctor of Philosophy (PhD)

College

Engineering

Department/School/Program

Computer Science

Faculty

Dr. Simone Silvestri

Abstract

Cyber-Physical Systems (CPS) represent a transformative paradigm that integrates sensing, computation, and actuation through networked systems to enable intelligent, context-aware applications across various domains. Despite their growing potential, CPS still face critical challenges, particularly in maintaining reliable, low-latency communication and efficient data processing in resource-constrained and dynamic environments. These challenges are further magnified in rural or remote deployments, where traditional network infrastructure is often unavailable or unreliable. This dissertation addresses these challenges through the development of optimization and inference algorithms that enhance network performance in Software-Defined Networks (SDN), using techniques such as reinforcement learning and network tomography for efficient routing. A central focus of this thesis is the application of CPS in smart agriculture, a domain with immense potential to benefit from CPS technologies to improve farming efficiency and sustainability, yet hindered by barriers such as limited broadband connectivity, harsh environmental conditions, power limitations, and low adoption rates among farmers. To overcome these challenges, this dissertation further presents contributions in three key areas: i) efficient data collection to bridge the connectivity gap in rural farms, ii) edge intelligence for resource-aware data processing, and iii) data-driven decision support to optimize farming resources, such as fertilizer and irrigation. In the context of agricultural sensor data collection, I propose two UAV-based data collection frameworks DRONE and CROP, designed for high-precision crop monitoring in less connected rural farms. These frameworks leverage machine learning and statistical inference, respectively, to minimize UAV energy consumption and operational costs while ensuring informative data acquisition and monitoring accuracy. To validate their effectiveness, the proposed methods are deployed on a drone testbed equipped with 27 UV irradiance sensors. Experimental evaluations demonstrate that the proposed methods significantly outperform existing approaches in terms of monitoring accuracy, energy efficiency, and UAV operational cost. Additionally, to enhance data processing in the absence of broadband connectivity, I propose iCrop+, a LoRa-based crop disease detection system that integrates on-device AI with deep learning for remote processing. iCrop+ intelligently offloads data via LoRa communication using a combination of category-based and sample-based offloading strategies to optimize accuracy and efficiency. A prototype of iCrop+ is developed and evaluated through both lab and open-field experiments, demonstrating its ability to detect crop diseases accurately while reducing detection latency and minimizing data transmission requirements. Finally, to close the CPS loop in agriculture, I introduce FertilizeSmart, a data-driven decision support tool that adaptively optimizes critical farming resources, such as nitrogen fertilization and irrigation, under dynamic weather conditions. This tool integrates domain knowledge with machine learning and simulation-based models to support sustainable and precise resource management.

Collectively, the contributions of this dissertation advance the network management and monitoring efficiency of SDN in CPS and CPS-empowered smart agriculture by enabling intelligent data collection, efficient edge processing, and sustainable decision-making. These innovations support more resilient and sustainable farming practices, particularly in less connected and resource-constrained agricultural environments.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2025.72

Funding Information

This dissertation was supported by the National Science Foundation Smart and Connected Community funded project “Smart Integrated Farm Network for Rural Agricultural Communities” (SIRAC), award number 1952045.

Download

Share

COinS