In recent years, robotic rehabilitation has gained significant traction in the recovering process of subjects affected by physical impairment. In particular, thanks to the application of compliant control strategies, it is possible to adapt the robot behaviour to the patient's needs by varying each joint stiffness. Ideally, the assistance provided by the exoskeleton should mimic the adaptive behaviour of a therapist during physical interaction with a patient. However, proper tuning of the exoskeleton settings is hindered by the challenge of mapping the patient's impairment level onto appropriate control parameters, thus limiting the robot adaptability. This work proposes the use of surface electromyography for the estimation of the variability of the therapist arm stiffness in order to improve the exoskeleton interaction. The therapist's muscle activity is used to estimate its arm endpoint stiffness, which is then mapped to the exoskeleton impedance parameters. The endpoint stiffness of the human arm has been inferred on 6 able-bodied subjects, demonstrating the ability of the proposed approach to estimate variations in a therapist's endpoint stiffness during patient guidance. Since the process aims at monitoring the stiffness variation, the profile generated for the exoskeleton tuning must be a time dependant signal. Consequently, the exoskeleton becomes a time-variant system, requiring explicit guarantees of stability. In order to overcome this challenge, this study proposes two different variable impedance control strategies: an offline optimization control law based on Lyapunov's theory to directly guarantee system stability and the port-based modelling approach known as energy tank to guarantee system passivity and therefore stable interaction with passive environments. The stability properties of both control strategies were first evaluated on a 1 degree of freedom system. The application of the proposed control strategies has been able to stabilize the system under conditions that would otherwise incur instability even in the absence of interaction. Finally, both control strategies were implemented and validated on AllyArm, a commercially available multi-degree of freedom rehabilitation exoskeleton for the upper-limb.
Negli ultimi anni, la riabilitazione robotica ha acquisito un ruolo sempre più rilevante nel processo di recupero di soggetti affetti da disabilità fisiche. In particolare, grazie a strategie di controllo complianti, è possibile adattare il comportamento del robot alle esigenze del paziente variando la rigidezza di ciascuna articolazione. Idealmente, l’assistenza fornita dall’esoscheletro dovrebbe imitare il comportamento adattivo di un terapista durante l’interazione fisica con il paziente. Tuttavia, la corretta regolazione dei parametri dell’esoscheletro è ostacolata dalla difficoltà nel tradurre il livello di compromissione funzionale del paziente in parametri di controllo, limitando così la capacità di adattamento del robot. Questo lavoro propone l’utilizzo dell’elettromiografia di superficie per stimare la variabilità della rigidezza del braccio del terapista, al fine di migliorare l’interazione dell’esoscheletro. L’attività muscolare del terapista viene utilizzata per stimare la rigidezza all’estremità operativa del braccio, che viene successivamente mappata sui parametri di impedenza dell’esoscheletro. Mediante la sEMG, la rigidezza del braccio umano è stata stimata su 6 soggetti sani, dimostrando la capacità dell’approccio proposto di rilevare le variazioni della rigidezza del terapista durante la guida del paziente. Poiché il processo mira a monitorare le variazioni di rigidezza, il profilo generato per la regolazione dell’esoscheletro deve essere un segnale dipendente dal tempo. Di conseguenza, l’esoscheletro diventa un sistema tempo-variante, richiedendo esplicite garanzie di stabilità. Per affrontare questa sfida, lo studio propone due diverse strategie di controllo a impedenza variabile: una legge di controllo offline ad ottimizzazione basata sulla teoria di Lyapunov, per garantire direttamente la stabilità del sistema, e un approccio noto come energy tank, che garantisce la passività del sistema e quindi un’interazione stabile con ambienti passivi. Le proprietà di stabilità di entrambe le strategie di controllo sono state inizialmente valutate su un sistema a un grado di libertà. L’applicazione delle strategie proposte ha dimostrato di essere in grado di stabilizzare il sistema in condizioni che avrebbero altrimenti causato instabilità. Infine, le strategie di controllo sono state implementate su AllyArm, un esoscheletro commerciale per la riabilitazione dell’arto superiore.
Estimating therapist endpoint stiffness for adaptive control of a rehabilitation exosekeleton
ZANELLA, FRANCESCO
2025/2026
Abstract
In recent years, robotic rehabilitation has gained significant traction in the recovering process of subjects affected by physical impairment. In particular, thanks to the application of compliant control strategies, it is possible to adapt the robot behaviour to the patient's needs by varying each joint stiffness. Ideally, the assistance provided by the exoskeleton should mimic the adaptive behaviour of a therapist during physical interaction with a patient. However, proper tuning of the exoskeleton settings is hindered by the challenge of mapping the patient's impairment level onto appropriate control parameters, thus limiting the robot adaptability. This work proposes the use of surface electromyography for the estimation of the variability of the therapist arm stiffness in order to improve the exoskeleton interaction. The therapist's muscle activity is used to estimate its arm endpoint stiffness, which is then mapped to the exoskeleton impedance parameters. The endpoint stiffness of the human arm has been inferred on 6 able-bodied subjects, demonstrating the ability of the proposed approach to estimate variations in a therapist's endpoint stiffness during patient guidance. Since the process aims at monitoring the stiffness variation, the profile generated for the exoskeleton tuning must be a time dependant signal. Consequently, the exoskeleton becomes a time-variant system, requiring explicit guarantees of stability. In order to overcome this challenge, this study proposes two different variable impedance control strategies: an offline optimization control law based on Lyapunov's theory to directly guarantee system stability and the port-based modelling approach known as energy tank to guarantee system passivity and therefore stable interaction with passive environments. The stability properties of both control strategies were first evaluated on a 1 degree of freedom system. The application of the proposed control strategies has been able to stabilize the system under conditions that would otherwise incur instability even in the absence of interaction. Finally, both control strategies were implemented and validated on AllyArm, a commercially available multi-degree of freedom rehabilitation exoskeleton for the upper-limb.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_07_Zanella_Executive_Summary.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
3.08 MB
Formato
Adobe PDF
|
3.08 MB | Adobe PDF | Visualizza/Apri |
|
2026_07_Zanella_Thesis.pdf
non accessibile
Descrizione: Master Thesis
Dimensione
38.05 MB
Formato
Adobe PDF
|
38.05 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/10589/261213