This Ph.D. dissertation will go through the development of enabling technologies that are inspired by the concepts envisioned by Industry 5.0. The research activity hereby described is aimed at improving the quality of life of industrial workers by means of wearable technologies and artificial intelligence. These research topics match with the core concepts of Industry 5.0, which are: wearable techs and robotics, human-robot collaboration, and technology evaluation and health monitoring. The exploration of such research topics – spanning across several disciplines of engineering, robotics, and computer science – led to the development of a set of modular yet interconnected technologies with the common goal of tackling work-related musculoskeletal disorders (WRMDs) for the benefit of millions of workers. Indeed, the end goal of implementing such technologies is to reduce the massive impact that work-related musculoskeletal disorders have on workers in the industrial sector. The lifetime prevalence of low-back pain is greater than 50 % for industrial workers. This means that hundreds of millions of people worldwide have been suffering from low-back pain. Existing low-back exoskeletons have the potential to empower workers, partially relieving the physical stress they experience on a daily basis. On the other hand, there are several critical aspects of this technology that remain unsolved, especially in terms of end-user acceptability, which limit its real-world adoption. The investigation of wearable tech and robotics resulted in the development of a low-back exoskeleton prototype. The study of human-robot collaboration led to the development of machine learning models for payload estimation and human activity recognition. Finally, real-time task ergonomics feedback was implemented in the context of technology evaluation. Furthermore, intersections among the core concepts of Industry 5.0 generated further research topics. The development of a compliant robotic actuator involved wearable tech and robotics and human-robot collaboration. The development of a protocol for exoskeleton validation regarded, again, wearable tech and robotics as well as technology evaluation. This PhD dissertation will go into the details of each of the research areas described above, highlighting the reached goals as well as the challenges and the open problems. The first topic (cf. Chapter 2) is a critical analysis of the state of the art, focusing on low-back assistance for industrial workers. This research shows how real-world adoption of occupational low-back exoskeletons is below expectations. As detailed in Chapter 2, this can be explained by the lack of rigorous standards for their performance evaluation, which is fundamental to understand their impact and compare available solutions. We will see how the lack of rigorous industrial standards may have been limiting exoskeleton adoption in the industrial scenario. Having depicted the starting point of this research, the next contribution (cf. Chapter 3) is focused on the development of an occupational low-back exoskeleton. An existing design featuring a peculiar backbone-tracking kinematic mechanism has been optimized for fused deposition molding. FDM 3D printing provides rapid manufacturing and cost efficiency, making it highly suitable for prototyping applications. The 3D-printed prototype can thus be exploited as a modular research platform. Exploring the concept of worker’s health monitoring, we have equipped the exoskeleton with an embedded real-time ergonomics feedback system. This system – developed as a stand-alone module – is based on a set of wearable inertial sensors that implement kinematic-based rules such as the RULA (Rapid Upper-Limb Assessment). Real-time feedback is then given to the wearer, enabling the exoskeleton to not only provide physical assistance but also improve task ergonomics. Chapter 3 also shows how 3D-printing can allow fast prototyping and scalable user-centric design. After the first iteration, the prototype has been evaluated both by healthy volunteers with a questionnaire and by potential end-users with the System Usability Scale (SUS): results just above the acceptability threshold are promising for further improvement in the following design iterations. The exploitation of FDM 3D-printing continued throughout research on the topics of wearable tech and robotics and human-robot collaboration. FDM prototyping has been investigated for the development of a compliant robotic actuator. Chapter 4 shows the design and characterization of a 3D-printed cycloidal speed reducer, then coupled to a BLDC motor to obtain an actuator for occupational exoskeletons. Using carbon-fiber reinforced plastic materials, we show the feasibility of a cost-effective design that allows compliant human-robot interaction. In the context of human-robot interaction, Chapter 5 shows the exploitation of artificial intelligence for the development of a high-level exoskeleton control system. In particular, we show how a Recurrent Neural Network (RNN) based on Long Short-Term Memory layers can learn to recognize human activities and estimate the lifted payload. This is achieved relying on time-series data measured by wearable inertial sensors only, in order not to jeopardize the wearability and the invasiveness of the exoskeleton. The final research topic connects to the initial assessment of the state of the art and concludes by proposing a methodology for metrics-based evaluation of occupational exoskeletons aimed to obtain rigorous and repeatable results. In the context of technology evaluation, we highlight the importance of using performance metrics to assess both primary (i.e., intended) and secondary (i.e., collateral) effects of the exoskeleton under evaluation. Finally, we show an example considering our passive low-back exoskeleton prototype and a manual material handling task. Reducing the impact on industrial workers of low-back pain and other work-related musculoskeletal disorders is an ambitious goal and its actual achievement may be hard to evaluate in the scope of a Ph.D. dissertation. Instead, we can evaluate several more specific secondary objectives. This can be done by looking at the achievements of this work, which are the developed enabling technologies, weighing their novelty against their limitations. In light of the above, this research project is not an end-point in the effort towards the reduction of the impact of WRMDs, yet it hopes to be a valid groundwork for forthcoming research aimed at reaching this goal.
Questa tesi di Dottorato affronta lo sviluppo di tecnologie definite enabling technologies che si ispirano ai concetti su cui si basa l’Industria 5.0. L’attività di ricerca descritta di seguito è finalizzata a migliorare la qualità di vita dei lavoratori impiegati nel settore industriale tramite tecnologie indossabili e intelligenza artificiale. I temi di ricerca qui trattati corrispondono ai concetti fondamentali dell’Industria 5.0, ovvero wearable tech, robotica, collaborazione uomo-robot, validazione della tecnologia e monitoraggio della salute. L’esplorazione di questi temi di ricerca – che abbracciano diverse discipline dell’ingegneria, della robotica e dell’informatica – ha portato allo sviluppo di una serie di tecnologie modulari ma interconnesse, aventi l’obiettivo comune di affrontare i disturbi muscoloscheletrici legati al lavoro (Work-Related Musculoskeletal Disorders, WRMD) a beneficio di milioni di lavoratori. L’obiettivo finale dell’implementazione di queste tecnologie è infatti quello di ridurre l’impatto, ad oggi considerevole, che i disturbi muscoloscheletrici legati all’attività lavorativa hanno sui lavoratori del settore industriale. La prevalenza nel corso della vita del dolore lombare è superiore al 50 % per i lavoratori di tale settore. Ciò significa che centinaia di milioni di persone in tutto il mondo soffrono di lombalgia e altri disturbi musculoscheletrici. Gli esoscheletri lombari esistenti hanno il potenziale di fornire assistenza ai lavoratori, alleviando in parte lo stress fisico che subiscono quotidianamente. D’altra parte, ci sono diversi aspetti critici di questa tecnologia che, ad oggi, rimangono irrisolti, soprattutto in termini di accettabilità da parte dell’utente finale, che ne limitano l’adozione in ambito lavorativo. Facendo riferimento all’illustrazione delle tematiche presentati in questa tesi, lo studio di tecnologie indossabili e robotica ha portato allo sviluppo di un prototipo di esoscheletro per la schiena. Lo studio della collaborazione uomo-robot ha portato allo sviluppo di modelli di machine learning per la stima del carico sollevato e per il riconoscimento delle attività svolte dal lavoratore. Infine, è stato implementato un feedback in tempo reale sull’ergonomia delle attività svolte nel contesto della validazione della tecnologia. Inoltre, le intersezioni tra i concetti fondamentali dell’Industria 5.0 hanno generato ulteriori argomenti di ricerca. Lo sviluppo di un attuatore robotico compliante coinvolge temi come le tecnologie indossabili, la robotica e la collaborazione uomo-robot. Lo sviluppo di un protocollo per la validazione di esoscheletri industriali coinvolge, ancora una volta, la tecnologia indossabile e la robotica, nonché la validazione della tecnologia. Questa tesi di dottorato entrerà nel dettaglio di ciascuna delle aree di ricerca sopra descritte, evidenziando gli obiettivi raggiunti, le sfide e i problemi aperti. Il primo argomento (cfr. Capitolo 2) è un’analisi critica dello stato dell’arte, incentrata sugli esoscheletri per i lavoratori industriali. Questa ricerca mostra come l’adozione in ambito lavorativo di esoscheletri occupazionali per la schiena sia inferiore alle aspettative. Come illustrato nel Capitolo 2, ciò si spiega con la mancanza di standard rigorosi per la valutazione degli esoscheletri stessi, fondamentale per comprenderne l’impatto e confrontare le soluzioni disponibili in termini numerici. Vedremo come la mancanza di standard industriali rigorosi possa aver limitato l’adozione degli esoscheletri nello scenario industriale. Dopo aver illustrato il punto di partenza di questa ricerca, il contributo successivo (cfr. Capitolo 3) è incentrato sullo sviluppo di un esoscheletro occupazionale per la schiena. Un progetto giá esistente, caratterizzato da un peculiare meccanismo cinematico in grado di tracciare l’allungamento della spina dorsale, é stato ottimizzato per la realizzazione di un prototipo mediante stampa 3D. Infatti, la stampa 3D con tecnologia FDM (Fused Mold Deposition) offre la possibilità di ottenere velocità del processo produttivo e un’elevata efficienza dei costi. Questa tecnologia è quindi molto adatta alle applicazioni di prototipazione e sviluppo di prototipi in ambito di ricerca. Il prototipo stampato in 3D può infatti essere sfruttato come piattaforma di ricerca modulare. Esplorando il concetto di monitoraggio della salute del lavoratore, l’esoscheletro è stato dotato di un sistema di feedback ergonomico in tempo reale. Questo sistema, sviluppato come modulo indipendente, si basa su una serie di sensori inerziali indossabili che implementano regole basate sulla cinematica, come la RULA (Rapid Upper-Limb Assessment). Il feedback in tempo reale viene quindi fornito all’utilizzatore, consentendo all’esoscheletro non solo di fornire assistenza fisica, ma anche di migliorare l’ergonomia delle singole attivitá lavorative. Il Capitolo 3 mostra anche come i vantaggi della prototipazione mediante stampa 3D possano essere applicati al concetto di user-centric design. Dopo la prima iterazione, infatti, il prototipo di escosheletro è stato valutato sia da volontari sani, mediante un questionario, sia da potenziali utenti finali, mediante la System Usability Scale (SUS): i risultati appena al di sopra della soglia di accettabilitá sono promettenti per ulteriori miglioramenti nelle successive iterazioni di progettazione. L’utilizzo della stampa 3D è proseguito nel corso della ricerca sui temi della tecnologia indossabile, della robotica e della collaborazione uomo-robot. La prototipazione FDM é stata utilizzata anche per lo sviluppo di un attuatore robotico compliante. Il Capitolo 4 mostra la progettazione e la caratterizzazione di un riduttore di velocitá cicloidale stampato in 3D, poi accoppiato a un motore BLDC. Il risultato è un attuatore per eso- scheletri occupazionali. Utilizzando materiali plastici rinforzati con fibra di carbonio, nel Capitolo dimostriamo la fattibilità di un progetto che consente di ottenere con costi ridotto un attuatore capace di garantire un’interazione uomo-robot efficace. Nel contesto dell’interazione uomo-robot, il Capitolo 5 mostra l’utilizzo dell’intelligenza artificiale per lo sviluppo di un sistema di controllo ad alto livello per esoscheletri occupazionali. In particolare, si mostra come l’utilizzo di Recurrent Neural Network (RNN) con architettura di tipo LSTM (long short-term memory) possa imparare a riconoscere le attività umane e a stimare il carico sollevato. Questo risultato è stato ottenuto basandosi solo su segnali misurati da sensori inerziali indossabili, per non compromettere l’indossabilità e l’invasività dell’esoscheletro. L’ultimo tema di ricerca si ricollega alla valutazione iniziale dello stato dell’arte e si conclude proponendo una metodologia per la valutazione basata su metriche e indicatori di performance per gli esoscheletri occupazionali. Tale metodologia è volta a ottenere risultati rigorosi e ripetibili. Nel contesto della valutazione tecnologica, sottolineiamo l’importanza di utilizzare metriche per valutare sia gli effetti primari (cioè quelli pre- visti) sia quelli secondari (cioè quelli collaterali) dell’esoscheletro in esame. Infine, mostriamo un esempio che prende in considerazione il nostro prototipo di esoscheletro passivo per la schiena e una attività di movimentazione manuale di carichi. Ridurre l’impatto del dolore lombare (low-back pain) e di altri disturbi muscoloscheletrici legati all’attività nell’ambito industriale è un obiettivo ambizioso e il suo effettivo raggiungimento può essere di difficile valutazione nell’ambito di una tesi di dottorato. Possiamo invece valutare diversi obiettivi secondari più specifici. Ciò può essere fatto esaminando i risultati di questo lavoro, che sono le enabling technologies sviluppate, soppesando la loro novità rispetto ai loro limiti. Alla luce di quanto detto, questo progetto di ricerca non rappresenta un punto di arrivo nello sforzo di ridurre l’impatto dei disturbi musculoscheletrici, ma spera di essere un valido punto di partenza per ricerche future volte a raggiungere questo obiettivo.
Enabling technologies towards industry 5.0: reducing the impact of musculoskeletal disorders exploiting wearables and artificial intelligence
Pesenti, Mattia
2023/2024
Abstract
This Ph.D. dissertation will go through the development of enabling technologies that are inspired by the concepts envisioned by Industry 5.0. The research activity hereby described is aimed at improving the quality of life of industrial workers by means of wearable technologies and artificial intelligence. These research topics match with the core concepts of Industry 5.0, which are: wearable techs and robotics, human-robot collaboration, and technology evaluation and health monitoring. The exploration of such research topics – spanning across several disciplines of engineering, robotics, and computer science – led to the development of a set of modular yet interconnected technologies with the common goal of tackling work-related musculoskeletal disorders (WRMDs) for the benefit of millions of workers. Indeed, the end goal of implementing such technologies is to reduce the massive impact that work-related musculoskeletal disorders have on workers in the industrial sector. The lifetime prevalence of low-back pain is greater than 50 % for industrial workers. This means that hundreds of millions of people worldwide have been suffering from low-back pain. Existing low-back exoskeletons have the potential to empower workers, partially relieving the physical stress they experience on a daily basis. On the other hand, there are several critical aspects of this technology that remain unsolved, especially in terms of end-user acceptability, which limit its real-world adoption. The investigation of wearable tech and robotics resulted in the development of a low-back exoskeleton prototype. The study of human-robot collaboration led to the development of machine learning models for payload estimation and human activity recognition. Finally, real-time task ergonomics feedback was implemented in the context of technology evaluation. Furthermore, intersections among the core concepts of Industry 5.0 generated further research topics. The development of a compliant robotic actuator involved wearable tech and robotics and human-robot collaboration. The development of a protocol for exoskeleton validation regarded, again, wearable tech and robotics as well as technology evaluation. This PhD dissertation will go into the details of each of the research areas described above, highlighting the reached goals as well as the challenges and the open problems. The first topic (cf. Chapter 2) is a critical analysis of the state of the art, focusing on low-back assistance for industrial workers. This research shows how real-world adoption of occupational low-back exoskeletons is below expectations. As detailed in Chapter 2, this can be explained by the lack of rigorous standards for their performance evaluation, which is fundamental to understand their impact and compare available solutions. We will see how the lack of rigorous industrial standards may have been limiting exoskeleton adoption in the industrial scenario. Having depicted the starting point of this research, the next contribution (cf. Chapter 3) is focused on the development of an occupational low-back exoskeleton. An existing design featuring a peculiar backbone-tracking kinematic mechanism has been optimized for fused deposition molding. FDM 3D printing provides rapid manufacturing and cost efficiency, making it highly suitable for prototyping applications. The 3D-printed prototype can thus be exploited as a modular research platform. Exploring the concept of worker’s health monitoring, we have equipped the exoskeleton with an embedded real-time ergonomics feedback system. This system – developed as a stand-alone module – is based on a set of wearable inertial sensors that implement kinematic-based rules such as the RULA (Rapid Upper-Limb Assessment). Real-time feedback is then given to the wearer, enabling the exoskeleton to not only provide physical assistance but also improve task ergonomics. Chapter 3 also shows how 3D-printing can allow fast prototyping and scalable user-centric design. After the first iteration, the prototype has been evaluated both by healthy volunteers with a questionnaire and by potential end-users with the System Usability Scale (SUS): results just above the acceptability threshold are promising for further improvement in the following design iterations. The exploitation of FDM 3D-printing continued throughout research on the topics of wearable tech and robotics and human-robot collaboration. FDM prototyping has been investigated for the development of a compliant robotic actuator. Chapter 4 shows the design and characterization of a 3D-printed cycloidal speed reducer, then coupled to a BLDC motor to obtain an actuator for occupational exoskeletons. Using carbon-fiber reinforced plastic materials, we show the feasibility of a cost-effective design that allows compliant human-robot interaction. In the context of human-robot interaction, Chapter 5 shows the exploitation of artificial intelligence for the development of a high-level exoskeleton control system. In particular, we show how a Recurrent Neural Network (RNN) based on Long Short-Term Memory layers can learn to recognize human activities and estimate the lifted payload. This is achieved relying on time-series data measured by wearable inertial sensors only, in order not to jeopardize the wearability and the invasiveness of the exoskeleton. The final research topic connects to the initial assessment of the state of the art and concludes by proposing a methodology for metrics-based evaluation of occupational exoskeletons aimed to obtain rigorous and repeatable results. In the context of technology evaluation, we highlight the importance of using performance metrics to assess both primary (i.e., intended) and secondary (i.e., collateral) effects of the exoskeleton under evaluation. Finally, we show an example considering our passive low-back exoskeleton prototype and a manual material handling task. Reducing the impact on industrial workers of low-back pain and other work-related musculoskeletal disorders is an ambitious goal and its actual achievement may be hard to evaluate in the scope of a Ph.D. dissertation. Instead, we can evaluate several more specific secondary objectives. This can be done by looking at the achievements of this work, which are the developed enabling technologies, weighing their novelty against their limitations. In light of the above, this research project is not an end-point in the effort towards the reduction of the impact of WRMDs, yet it hopes to be a valid groundwork for forthcoming research aimed at reaching this goal.File | Dimensione | Formato | |
---|---|---|---|
Pesenti_Mattia_PhD_thesis_2024.pdf
non accessibile
Descrizione: ENABLING TECHNOLOGIES TOWARDS INDUSTRY 5.0: REDUCING THE IMPACT OF MUSCULOSKELETAL DISORDERS EXPLOITING WEARABLES AND ARTIFICIAL INTELLIGENCE
Dimensione
11.02 MB
Formato
Adobe PDF
|
11.02 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/228552