Lower-limb exoskeletons (LLEs) represent one of the most promising technologies for enhancing gait rehabilitation in individuals with spinal cord injuries or stroke. Rehabilitation with LLEs has shown strong effectiveness in providing intensive, task-specific, and repetitive training while enhancing user engagement and independence. Despite the numerous advantages associated with LLE-based rehabilitation, substantial challenges persist in LLEs mechanical design and control architectures. A primary challenge lies in motion reproduction. In this context, achieving human-like behavior, ensuring safe and adaptive motion trajectories, and delivering energy-efficient joint actuation remain critical research issues. Another key challenge is to increase the number of active degrees of freedom (DoFs) without compromising device mass, overall performance, or exoskeleton autonomy. Parallel to these efforts, the development of advanced assistive control strategies that enable seamless and efficient human–robot interaction remains a major area of investigation. The scientific objective of this doctoral thesis is to address these key challenges by developing and evaluating practical solutions through the analysis, implementation, and experimental validation of novel path-planning algorithms, active joint designs, and compliant control strategies, to advance the next generation of LLEs. The first aim of this thesis is to develop a trajectory planning method capable of generating joint trajectories for LLEs that simultaneously combine human-like and flexible gait behavior with feasible and energy-efficient joint motions. In this regard, a novel motion-planning framework was developed by combining spatial gait modularity with joint-space optimization, employing Bézier curve parameterization and multi-joint optimization methods. This framework was implemented on the TwinMed LLE and experimentally validated with healthy participants. The proposed approach enabled a wide range of feasible gait patterns while maintaining a high degree of similarity to natural human gait. Moreover, experimental evaluations with healthy users revealed that joint power requirements during walking were substantially reduced, resulting in smoother and more comfortable movement. As a result, overall energy efficiency improved, extending the device’s effective battery life and allowing for longer rehabilitation sessions. The achieved results represent a step toward a unified LLE trajectory-generation framework that integrates human-likeness and spatial modularity with energy efficiency and enhanced device autonomy. The second aim of this dissertation is the analysis and integration of a novel Active Ankle–Foot Orthosis (AAFO) into the TwinMed exoskeleton. To this end, an AAFO was designed for efficient integration into the LLE, providing user benefits without adversely affecting overall system performance. Specifically, the AAFO’s design and control aim to promote energy-efficient behavior by reducing the active power required to move the joint throughout the gait cycle. In its initial development phase, the joint prototype was validated through benchtop experiments, demonstrating its ability to track humanlike reference trajectories. Subsequently, the AAFO was integrated into the TwinMed exoskeleton’s control architecture. This integration incorporates a dedicated trajectory planning algorithm, enabling coordinated motion of the AAFO with the hip and knee joints. The proposed planning algorithm facilitates ground contact during the stance phase and enhances gait safety during the swing phase by increasing toe clearance. The resulting 6-DoFs walking pattern was experimentally evaluated with healthy participants. Findings indicate that the proposed AAFO effectively reduces the likelihood of stumbling during level-ground walking. Moreover, its integration into the LLE does not adversely affect the device’s efficiency or significantly increase power consumption. Furthermore, this dissertation aims to develop a novel assist-as-needed (AAN) control algorithm that ensures greater user freedom and a less constrained human–exoskeleton interaction. In this context, a tunnel-based force-field AAN controller was designed to deliver adaptive assistance while maintaining user autonomy. The proposed method relies on a time-independent control scheme that assists the user’s movements while correcting deviations from the target trajectory by applying forces at the end-effector, thereby emulating the guidance typically provided by a physiotherapist. The main innovation lies in the modular definition of the assistance tunnel, which enables varying levels of guidance across different phases of the gait cycle. The proposed control strategy was preliminarily validated on the TwinMed exoskeleton through benchtop experiments, demonstrating the feasibility of this novel approach for achieving modular and more compliant control in wearable robotic systems. In summary, this doctoral dissertation addresses several key research challenges in the field of LLEs by proposing practical solutions that span optimization-based path-planning algorithms, energy-efficient active joint integration, and control strategies for improved exoskeleton assistance. The work aims to make a significant contribution to LLE control toward a more efficient, user-centered robot-assisted rehabilitation.
Gli esoscheletri per gli arti inferiori (LLE) rappresentano una delle tecnologie più promettenti per migliorare la riabilitazione del cammino nelle persone con lesioni del midollo spinale o ictus. La riabilitazione con LLE ha dimostrato una forte efficacia nel fornire un allenamento intensivo, specifico per il compito e ripetitivo, aumentando al contempo il coinvolgimento dell’utente e il suo livello di autonomia. Nonostante i numerosi vantaggi associati alla riabilitazione basata su LLE, permangono tuttora sfide significative nella progettazione meccanica e nello sviluppo di architetture di controllo. Una delle principali criticità riguarda la riproduzione del movimento. In questo contesto, ottenere un comportamento simile a quello umano, garantire traiettorie di movimento sicure e adattive e fornire un’attuazione energeticamente efficiente restano questioni di ricerca fondamentali. Un’altra sfida rilevante consiste nell’aumentare il numero di gradi di libertà (DoF) attivi senza compromettere la massa del dispositivo, le prestazioni complessive o l’autonomia dell’esoscheletro. Parallelamente a questi sforzi, lo sviluppo di strategie avanzate di controllo assistivo che consentano un’interazione uomo–robot fluida ed efficiente rimane un importante ambito di ricerca. L’obiettivo scientifico di questa tesi di dottorato è affrontare queste sfide chiave sviluppando e valutando soluzioni pratiche attraverso l’analisi, l’implementazione e la validazione sperimentale di nuovi algoritmi di pianificazione del movimento, nuovi design di giunti attivi e nuove strategie di controllo complianti, al fine di promuovere lo sviluppo della prossima generazione di esoscheletri per gli arti inferiori. Il primo obiettivo di questa tesi è sviluppare un metodo di pianificazione delle traiettorie in grado di generare traiettorie articolari per esoscheletri di arto inferiore che combinino simultaneamente un comportamento del cammino simile a quello umano e modulabile con movimenti articolari fattibili ed energeticamente efficienti. A tal fine, è stato sviluppato un nuovo framework di pianificazione del movimento combinando la modularità spaziale del cammino con l’ottimizzazione nello spazio dei giunti, utilizzando la parametrizzazione mediante curve di Bézier e metodi di ottimizzazione multigiunto. Questo framework è stato implementato sull’esoscheletro TwinMed e validato sperimentalmente con partecipanti sani. L’approccio proposto ha permesso di ottenere un’ampia gamma di traiettorie di cammino, mantenendo un alto grado di somiglianza con il cammino umano naturale. Inoltre, le valutazioni sperimentali con utenti sani hanno evidenziato che i requisiti di potenza richiesti ai motori durante la deambulazione sono stati notevolmente ridotti, determinando un movimento più fluido e confortevole. Di conseguenza, è migliorata l’efficienza energetica complessiva, prolungando la durata effettiva della batteria del dispositivo e consentendo sessioni di riabilitazione più lunghe. I risultati ottenuti rappresentano un passo verso un framework unificato di generazione delle traiettorie per LLE che integri camminate più umane e modulabili con efficienza energetica e maggiore autonomia del dispositivo. Il secondo obiettivo di questa dissertazione si concentra sull’analisi e sull’integrazione di una nuova Ortesi Attiva di Caviglia (AAFO) nell’esoscheletro TwinMed. A tal fine, l’AAFO è stato progettato per un’integrazione efficiente nell’LLE, fornendo benefici all’utente senza compromettere le prestazioni complessive del sistema. In particolare, il progetto e il controllo dell’AAFO mirano a promuovere un comportamento energeticamente efficiente, riducendo la potenza attiva necessaria a muovere l’articolazione durante l’intero ciclo del cammino. Nella fase iniziale di sviluppo, il prototipo dell’AAFO è stato validato tramite esperimenti su banco, dimostrando la sua capacità di seguire traiettorie di riferimento simili a quelle umane. Successivamente, l’AAFO è stato integrato nell’architettura di controllo dell’esoscheletro TwinMed. Questa integrazione include un algoritmo dedicato di pianificazione delle traiettorie, che consente un movimento coordinato della caviglia con le articolazioni dell’anca e del ginocchio. L’algoritmo di pianificazione proposto facilita il contatto con il suolo durante la fase di appoggio e migliora la sicurezza del cammino durante la fase di swing, aumentando la toe clearance. Lo schema di cammino a 6 gradi di libertà risultante è stato sperimentalmente valutato con partecipanti sani. I risultati indicano che l’AAFO proposto riduce efficacemente il rischio di inciampo durante il cammino. Inoltre, la sua integrazione nell’LLE non compromette l’efficienza del dispositivo né aumenta significativamente il consumo energetico. Inoltre, questa dissertazione si propone di sviluppare un nuovo algoritmo di controllo assist-as-needed (AAN) in grado di garantire maggiore libertà all’utente e un’interazione uomo–esoscheletro meno vincolata. In questo contesto, è stato progettato un controllore AAN basato su un tunnel di forza, in grado di fornire un’assistenza adattiva pur mantenendo l’autonomia dell’utente. Il metodo proposto si basa su uno schema di controllo indipendente dal tempo, che supporta i movimenti dell’utente correggendo le deviazioni dalla traiettoria target tramite l’applicazione di forze all’end-effector, emulando così la guida tipicamente fornita da un fisioterapista. L’innovazione principale risiede nella definizione modulare del tunnel di assistenza, che consente di variare i livelli di guida nelle diverse fasi del ciclo del cammino. La strategia di controllo proposta è stata validata preliminarmente sull’esoscheletro TwinMed tramite esperimenti su banco, dimostrando la fattibilità di questo nuovo approccio per ottenere un controllo modulare e più cedevole nei sistemi robotici indossabili. In sintesi, questa tesi di dottorato affronta diverse sfide di ricerca chiave nel campo degli esoscheletri di arto inferiore, proponendo soluzioni pratiche che spaziano dagli algoritmi di pianificazione del percorso basati su ottimizzazione, all’integrazione di articolazioni attive energeticamente efficienti, fino alle strategie di controllo per migliorare l’assistenza fornita dall’esoscheletro. Il lavoro mira a fornire un contributo significativo al controllo degli esoscheletri, orientato a una riabilitazione assistita da robot più efficiente e incentrata sull’utente.
Optimizing motion planning, design, and control for rehabilitative Lower Limb Exoskeletons
Giannattasio, Raffaele
2025/2026
Abstract
Lower-limb exoskeletons (LLEs) represent one of the most promising technologies for enhancing gait rehabilitation in individuals with spinal cord injuries or stroke. Rehabilitation with LLEs has shown strong effectiveness in providing intensive, task-specific, and repetitive training while enhancing user engagement and independence. Despite the numerous advantages associated with LLE-based rehabilitation, substantial challenges persist in LLEs mechanical design and control architectures. A primary challenge lies in motion reproduction. In this context, achieving human-like behavior, ensuring safe and adaptive motion trajectories, and delivering energy-efficient joint actuation remain critical research issues. Another key challenge is to increase the number of active degrees of freedom (DoFs) without compromising device mass, overall performance, or exoskeleton autonomy. Parallel to these efforts, the development of advanced assistive control strategies that enable seamless and efficient human–robot interaction remains a major area of investigation. The scientific objective of this doctoral thesis is to address these key challenges by developing and evaluating practical solutions through the analysis, implementation, and experimental validation of novel path-planning algorithms, active joint designs, and compliant control strategies, to advance the next generation of LLEs. The first aim of this thesis is to develop a trajectory planning method capable of generating joint trajectories for LLEs that simultaneously combine human-like and flexible gait behavior with feasible and energy-efficient joint motions. In this regard, a novel motion-planning framework was developed by combining spatial gait modularity with joint-space optimization, employing Bézier curve parameterization and multi-joint optimization methods. This framework was implemented on the TwinMed LLE and experimentally validated with healthy participants. The proposed approach enabled a wide range of feasible gait patterns while maintaining a high degree of similarity to natural human gait. Moreover, experimental evaluations with healthy users revealed that joint power requirements during walking were substantially reduced, resulting in smoother and more comfortable movement. As a result, overall energy efficiency improved, extending the device’s effective battery life and allowing for longer rehabilitation sessions. The achieved results represent a step toward a unified LLE trajectory-generation framework that integrates human-likeness and spatial modularity with energy efficiency and enhanced device autonomy. The second aim of this dissertation is the analysis and integration of a novel Active Ankle–Foot Orthosis (AAFO) into the TwinMed exoskeleton. To this end, an AAFO was designed for efficient integration into the LLE, providing user benefits without adversely affecting overall system performance. Specifically, the AAFO’s design and control aim to promote energy-efficient behavior by reducing the active power required to move the joint throughout the gait cycle. In its initial development phase, the joint prototype was validated through benchtop experiments, demonstrating its ability to track humanlike reference trajectories. Subsequently, the AAFO was integrated into the TwinMed exoskeleton’s control architecture. This integration incorporates a dedicated trajectory planning algorithm, enabling coordinated motion of the AAFO with the hip and knee joints. The proposed planning algorithm facilitates ground contact during the stance phase and enhances gait safety during the swing phase by increasing toe clearance. The resulting 6-DoFs walking pattern was experimentally evaluated with healthy participants. Findings indicate that the proposed AAFO effectively reduces the likelihood of stumbling during level-ground walking. Moreover, its integration into the LLE does not adversely affect the device’s efficiency or significantly increase power consumption. Furthermore, this dissertation aims to develop a novel assist-as-needed (AAN) control algorithm that ensures greater user freedom and a less constrained human–exoskeleton interaction. In this context, a tunnel-based force-field AAN controller was designed to deliver adaptive assistance while maintaining user autonomy. The proposed method relies on a time-independent control scheme that assists the user’s movements while correcting deviations from the target trajectory by applying forces at the end-effector, thereby emulating the guidance typically provided by a physiotherapist. The main innovation lies in the modular definition of the assistance tunnel, which enables varying levels of guidance across different phases of the gait cycle. The proposed control strategy was preliminarily validated on the TwinMed exoskeleton through benchtop experiments, demonstrating the feasibility of this novel approach for achieving modular and more compliant control in wearable robotic systems. In summary, this doctoral dissertation addresses several key research challenges in the field of LLEs by proposing practical solutions that span optimization-based path-planning algorithms, energy-efficient active joint integration, and control strategies for improved exoskeleton assistance. The work aims to make a significant contribution to LLE control toward a more efficient, user-centered robot-assisted rehabilitation.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_Giannattasio_phd.pdf
accessibile in internet per tutti a partire dal 01/05/2029
Dimensione
58.92 MB
Formato
Adobe PDF
|
58.92 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/10589/257077