A set of future trends (e.g., the global population growth, climate changing) and challenges targeting the agri-food system will force agriculture, and, thus, agricultural processes, to become more productive, sustainable, and smarter. Key enablers in supporting this trend of transformation, usually denoted as Smart Agriculture (SA) and Smart Farming (SF), will be a heterogeneous set of digital technologies, including drones, autonomous tractors, Artificial Intelligence (AI)-based algorithms, and Internet of Things (IoT). Among them, this last technology will play a crucial role in allowing the effective integration of all the others within the same system, able to automatically manage the productive activities of future high-tech farms (or smart farms). Indeed, IoT allows to build systems (namely, IoT/SA-oriented systems) able to gather, process, share, and store data relevant for agricultural processes, thus, enabling the deployment of Farm Management Systems (FMSs) which can create new value from farm-related data. In details, agricultural relevant data, coming from physical devices installed in the farms (e.g., sensors, tractors, smart collars) or external data sources (e.g., historical databases), are managed by a networking and computing (IoT) infrastructure, which processing components are distributed between local networks of devices placed in the farms and the Cloud (in other words, which implements both the edge computing and the Cloud computing paradigms). The development of the above mentioned IoT architecture has to deal with several challenges related to both IoT technologies (e.g., heterogeneity, modularity, scalability, and standardization) and agricultural scenarios (e.g., connectivity issues, harsh environment, and device power management). Despite several IoT/SA-oriented systems proposed in literature have significantly enhanced the management of sets of farm’s activities, there is still the need to define methodologies to (i) properly handle all the challenges characterizing IoT/SA-oriented systems, and to (ii) develop general-purpose IoT systems targeting SA applications. Consequently, the aim of this doctoral thesis is to cover these research gaps, taking the following four stages. First, the main aspects, challenges, technologies, and requirements to consider while designing and developing IoT systems targeting SA applications are identified and discussed. Second, a theoretical architectural model (or framework) to follow while developing general-purpose IoT/SA-oriented architectures, which are modular, scalable, able to easily integrate heterogeneous technologies, and accelerating the diffusion of smart farms is defined. This considering the theoretical and practical knowledge acquired from the literature review performed in the first stage and the on-field experience of this doctoral thesis’s author. Third, the framework proposed in the second stage is experimentally validated. This presenting (i) how two real IoT/SA-oriented systems have been developed following the presented architectural model; and (ii) the experimental results correlated to the two described systems. Fourth, some insights concerning the integration of AI tools (as NNs) within the proposed framework to develop data mining, analysis, and forecasting algorithms, enriching IoT/SA-oriented systems with intelligent capabilities, are provided. This describing the methodologies adopted to develop two forecasting NN-based algorithms, starting from data coming from one of the systems presented in the third stage and aiming to enhance the management of a greenhouse.
Un insieme di trend futuri (come l’aumento della popolazione mondiale e il cambiamento climatico) e sfide che il sistema agroalimentare dovrà affrontare, costringeranno l’agricoltura e, quindi, i suoi processi produttivi, a diventare più produttivi, sostenibili e intelligenti. Fattori chiave per abilitare la trasformazione del settore agricolo, un trend noto con termini quali Smart Agricolture (SA) e Smart Farming (SF), saranno un insieme di tecnologie digitali molto eterogenee, che includeranno droni, trattori a guida autonomi, algoritmi basati sull’intelligenza artificiale (IA), e Internet of Things (IoT). Tra queste, quest’ultima tecnologia svolgerà un ruolo cruciale nell’integrare efficacemente tutte le altre tecnologie all’interno di un unico sistema, in grado di gestire automaticamente le attività produttive di una futura fattoria altamente tecnologica (o smart farm). L’IoT consente infatti di costruire sistemi (IoT/SA-oriented systems) in grado di raccogliere, elaborare, condividere e memorizzare dati rilevanti per i processi agricoli, consentendo così l’implementazione di Farm Management Systems (FMS) in grado di creare nuovo valore dai dati di interesse. In dettaglio, i dati raccolti, provenienti da dispositivi fisici installati nelle fattorie (es. sensori, trattori, smart collar) o da fonti esterne al sistema (es. database storici), sono gestiti da un’infrastruttura di comunicazione e di calcolo (IoT), i quali componenti di elaborazione sono distribuiti tra reti locali di dispositivi installate nelle fattorie e il Cloud (in altre parole, questa infrastruttura implementa il paradigma dell’edge computing e del Cloud computing). Lo sviluppo della suddetta architettura basata su tecnologie IoT deve affrontare diverse sfide. Alcune di essi sono tipiche delle tecnologie IoT (eincludono eterogeneità, modularità, scalabilità e standardizzazione), mentre altre diventano più incisive quando sono coinvolti scenari agricoli (ad esempio, problemi di connettività, ambiente ostile e gestione dell’alimentazione dei dispositivi). Nonostante diversi sistemi IoT/SA-oriented proposti in letteratura abbiano notevolmente migliorato la gestione di alcune attività agricole, permane la necessità di definire metodologie per (i) gestire adeguatamente tutte le sfide che caratterizzano i sistemi IoT/SA-oriented e per (ii) sviluppare sistemi IoT general-purpose destinati ad applicazioni SA. Per questo motivo, l’obiettivo di questa tesi di dottorato è quello di coprire questi research gaps secondo le seguenti quattro fasi. Innanzitutto, sono identificati e discussi i principali aspetti, le sfide, le tecnologie e i requisiti da considerare durante la progettazione e lo sviluppo di sistemi IoT destinati ad applicazioni di SA. In secondo luogo, è definito un modello architetturale di riferimento (o framework) da seguire durante lo sviluppo di sistemi IoT/SA-oriented general-purpose, che siano modulari, scalabili, in grado di integrare facilmente tecnologie eterogenee e di accelerare la diffusione delle smart farm. Ciò considerando le conoscenze teoriche e pratiche acquisite durante l’approfondita analisi della letteratura effettuata nella prima fase e l’esperienza pratica maturata dall’autore di questa tesi di dottorato sul campo. Terzo, il framework proposto nella seconda fase è validato sperimentalmente. Questo presentando (i) come sono stati sviluppati due veri sistemi IoT/SA-oriented seguendo il modello architetturale presentato; e (ii) i risultati sperimentali correlati ai due sistemi descritti. In quarto luogo, vengono forniti alcuni approfondimenti riguardanti l’integrazione di strumenti di intelligenza artificiale (come le Neural Networks, NNs) nel framework proposto al fine di sviluppare algoritmi di data mining, analisi e previsione. Quest’ultimo obiettivo è raggiunto presentando due metodologie da adottate per sviluppare due algoritmi previsionali basati su NN, partendo dai dati raccolti mediante uno dei due sistemi presentati nella terza fase e con l’obiettivo di migliorare la gestione di una serra.
Design and Development of IoT Architectures for Smart Agriculture Systems
Gaia, Codeluppi;
2022
Abstract
A set of future trends (e.g., the global population growth, climate changing) and challenges targeting the agri-food system will force agriculture, and, thus, agricultural processes, to become more productive, sustainable, and smarter. Key enablers in supporting this trend of transformation, usually denoted as Smart Agriculture (SA) and Smart Farming (SF), will be a heterogeneous set of digital technologies, including drones, autonomous tractors, Artificial Intelligence (AI)-based algorithms, and Internet of Things (IoT). Among them, this last technology will play a crucial role in allowing the effective integration of all the others within the same system, able to automatically manage the productive activities of future high-tech farms (or smart farms). Indeed, IoT allows to build systems (namely, IoT/SA-oriented systems) able to gather, process, share, and store data relevant for agricultural processes, thus, enabling the deployment of Farm Management Systems (FMSs) which can create new value from farm-related data. In details, agricultural relevant data, coming from physical devices installed in the farms (e.g., sensors, tractors, smart collars) or external data sources (e.g., historical databases), are managed by a networking and computing (IoT) infrastructure, which processing components are distributed between local networks of devices placed in the farms and the Cloud (in other words, which implements both the edge computing and the Cloud computing paradigms). The development of the above mentioned IoT architecture has to deal with several challenges related to both IoT technologies (e.g., heterogeneity, modularity, scalability, and standardization) and agricultural scenarios (e.g., connectivity issues, harsh environment, and device power management). Despite several IoT/SA-oriented systems proposed in literature have significantly enhanced the management of sets of farm’s activities, there is still the need to define methodologies to (i) properly handle all the challenges characterizing IoT/SA-oriented systems, and to (ii) develop general-purpose IoT systems targeting SA applications. Consequently, the aim of this doctoral thesis is to cover these research gaps, taking the following four stages. First, the main aspects, challenges, technologies, and requirements to consider while designing and developing IoT systems targeting SA applications are identified and discussed. Second, a theoretical architectural model (or framework) to follow while developing general-purpose IoT/SA-oriented architectures, which are modular, scalable, able to easily integrate heterogeneous technologies, and accelerating the diffusion of smart farms is defined. This considering the theoretical and practical knowledge acquired from the literature review performed in the first stage and the on-field experience of this doctoral thesis’s author. Third, the framework proposed in the second stage is experimentally validated. This presenting (i) how two real IoT/SA-oriented systems have been developed following the presented architectural model; and (ii) the experimental results correlated to the two described systems. Fourth, some insights concerning the integration of AI tools (as NNs) within the proposed framework to develop data mining, analysis, and forecasting algorithms, enriching IoT/SA-oriented systems with intelligent capabilities, are provided. This describing the methodologies adopted to develop two forecasting NN-based algorithms, starting from data coming from one of the systems presented in the third stage and aiming to enhance the management of a greenhouse.| File | Dimensione | Formato | |
|---|---|---|---|
|
21_PhD_activites_report_Gaia_Codeluppi.pdf
non disponibili
Licenza:
Tutti i diritti riservati
Dimensione
57.92 kB
Formato
Adobe PDF
|
57.92 kB | Adobe PDF | |
|
21_PhD_thesis_official_rev_GaiaCodeluppi.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
91.02 MB
Formato
Adobe PDF
|
91.02 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/193164
URN:NBN:IT:UNIPR-193164