In populated environments, autonomous systems must demonstrate robust and accurate capabilities to account for the inherent unpredictability of human behavior. A typical example is BUDD-e, a guide robot designed to assist visually impaired users in navigating complex environments through a consistent pulling force provided by a Smart Tether system. This robot must be socially aware and capable of predicting and adapting to human movements to prevent discomfort or collisions. Consequently, this thesis presents novel real-time estimation methodologies tailored for assistive robotics operating in populated environments. In the first part a Kalman Filter-based approach is developed for the joint estimation of the distance and relative velocity of a user with respect to the BUDD-e robotic guide. Unlike the currently employed Target Tracking algorithm based on LiDAR measurements, the proposed system leverages measures obtained from a Smart Tether system, enabling robust performance even in case of sensor failures. The second part introduces a novel algorithm for the prompt detection of user stops inspired by multivariate statistical analysis. By analyzing the force and angular velocity applied by the user on the Smart Tether system, a velocity-dependent threshold-based strategy is proposed to identify user-initiated stops. The method is optimized to minimize detection delay and false positives, thus improving user safety during sudden decelerations. The final part focuses on predicting human trajectories in dynamic environments using a data-driven approach based on Koopman Operator theory. This approach enables the modeling of the nonlinear pedestrian behavior, allowing real-time trajectory prediction and adaptive navigation planning for autonomous robots in structured environments such as hospitals. The effectiveness of the proposed algorithms is demonstrated through extensive field testing, highlighting their advantages in terms of robustness, responsiveness, and integration in real-world applications.
In ambienti popolati, i sistemi autonomi devono dimostrarsi robusti e accurati per gestire l’imprevedibilità del comportamento umano. Un tipico esempio è BUDD-e, un robot guida progettato per assistere utenti ipovedenti nella navigazione di ambienti complessi attraverso una forza di trazione costante fornita da un sistema Smart Tether. Questo robot deve essere socialmente consapevole e in grado di prevedere e adattarsi ai movimenti umani per evitare disagi o collisioni. Di conseguenza, questa tesi presenta metodologie innovative di stima e controllo in tempo reale destinate a robot assistivi operanti in ambienti popolati. Nella prima parte viene sviluppato un approccio basato su un filtro di Kalman per la stima congiunta della distanza e della velocità relativa di un utente rispetto a BUDD-e. A differenza dell’algoritmo di tracking correntemente implementato e basato sulle misurazioni del LiDAR, l’approccio proposto utilizza dati di fase e velocità angolare del verricello provenienti dallo Smart Tether system, garantendo prestazioni robuste anche in caso di fallimento. La seconda parte introduce un algoritmo di rilevamento di frenate effettuate dall’utente, ispirato all’analisi statistica multivariata. Analizzando la forza e la velocità angolare applicate dall’utente sullo Smart Tether system, viene proposta una strategia basata su soglie dipendenti dalla velocità per identificare le frenate. Il metodo è ottimizzato per minimizzare il ritardo di rilevamento e i falsi positivi, migliorando così la sicurezza dell’utente durante decelerazioni improvvise. L’ultima parte è focalizzata sulla previsione delle traiettorie seguite da pedoni in ambienti dinamici mediante un approccio basato sulla teoria dell’Operatore di Koopman in grado di sfruttare dati sperimentali ottenuti dal sistema. Questo approccio consente di modellare il comportamento non lineare dei pedoni, permettendo la previsione in tempo reale delle traiettorie e una pianificazione adattiva della navigazione per robot autonomi. L’efficacia degli algoritmi proposti è verificata tramite test sperimentali, evidenziandone i vantaggi in termini di robustezza, reattività e integrazione in scenari reali.
Development and application of advanced estimation algorithms for assistive robotics in populated environments
Porfiri, Pierandrea;PIZZOCHERI, DAVIDE
2024/2025
Abstract
In populated environments, autonomous systems must demonstrate robust and accurate capabilities to account for the inherent unpredictability of human behavior. A typical example is BUDD-e, a guide robot designed to assist visually impaired users in navigating complex environments through a consistent pulling force provided by a Smart Tether system. This robot must be socially aware and capable of predicting and adapting to human movements to prevent discomfort or collisions. Consequently, this thesis presents novel real-time estimation methodologies tailored for assistive robotics operating in populated environments. In the first part a Kalman Filter-based approach is developed for the joint estimation of the distance and relative velocity of a user with respect to the BUDD-e robotic guide. Unlike the currently employed Target Tracking algorithm based on LiDAR measurements, the proposed system leverages measures obtained from a Smart Tether system, enabling robust performance even in case of sensor failures. The second part introduces a novel algorithm for the prompt detection of user stops inspired by multivariate statistical analysis. By analyzing the force and angular velocity applied by the user on the Smart Tether system, a velocity-dependent threshold-based strategy is proposed to identify user-initiated stops. The method is optimized to minimize detection delay and false positives, thus improving user safety during sudden decelerations. The final part focuses on predicting human trajectories in dynamic environments using a data-driven approach based on Koopman Operator theory. This approach enables the modeling of the nonlinear pedestrian behavior, allowing real-time trajectory prediction and adaptive navigation planning for autonomous robots in structured environments such as hospitals. The effectiveness of the proposed algorithms is demonstrated through extensive field testing, highlighting their advantages in terms of robustness, responsiveness, and integration in real-world applications.File | Dimensione | Formato | |
---|---|---|---|
2025_07_Pizzocheri_Porfiri_Thesis.pdf
accessibile in internet per tutti
Descrizione: Testo della tesi
Dimensione
7.27 MB
Formato
Adobe PDF
|
7.27 MB | Adobe PDF | Visualizza/Apri |
2025_07_Pizzocheri_Porfiri_ExecutiveSummary.pdf
accessibile in internet per tutti
Descrizione: Executive Summary
Dimensione
1.24 MB
Formato
Adobe PDF
|
1.24 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/10589/240522