The integration of robotic systems into environments shared with people requires that robots are not only safe and functional, but also socially acceptable. Social acceptance is defined as the extent to which users feel safe, understood, and comfortable when interacting with a robot, and depends on factors such as trust, transparency, compliance with social norms, and adaptability. This thesis addresses the topic by first proposing a reference model based on three components: definitions, requirements, and architecture. The requirements include both functional and non-functional aspects, such as usability, perceived usefulness, safety, privacy, and cultural alignment. From an architectural perspective, a socially acceptable robot must integrate multimodal perception, awareness of the surrounding context, safety, and adaptive control. Within this model, the research presented in this work focuses on two main aspects: user understanding and social navigation. User understanding concerns the ability of the robot to detect and interpret cognitive and emotional states, in order to personalize interaction and enhance user comfort and trust. To this end, classification methods based on multimodal sensors and biosignal analysis (EEG, ECG, GSR, facial expressions) were developed to estimate mental workload and stress levels in human-robot collaboration scenarios. The studies confirmed the validity of physiological indices as predictors of cognitive and emotional states and led to the creation of a public multimodal dataset dedicated to collaborative robotics. These tools allow the robot to dynamically adapt operational parameters, thereby improving the quality of interaction. Social navigation, which is crucial for mobile robots operating in dynamic and populated environments, was addressed through the development of a control framework capable of integrating spatial and temporal constraints to ensure socially acceptable behavior. The approach combines signal temporal logic (STL) with control barrier functions (CBFs), guaranteeing compliance with specifications and safety of interaction even under uncertainty and in the presence of dynamic obstacles. The framework was extended with additional constraints to maintain the user within the robot’s field of view and with robust formulations that incorporate filtering techniques, ensuring reliability even in complex real-world scenarios. In parallel, a cognitive architecture was proposed that integrates large language models (LLMs) to interpret heterogeneous inputs (emotional state, intentions, social context) and translate them into navigation parameters. This solution enables the robot to dynamically and contextually adapt its movement strategy, leading to the concept of intelligent social navigation. In addition, a lightweight architecture was developed for social guidance tasks, combining facial recognition, gesture detection, and skeleton tracking, in order to guarantee transparent and reactive behavior in public environments. Overall, the thesis demonstrates that social acceptance is not a static property of the robot, but an emergent phenomenon resulting from interaction with the user. The proposed solutions for user understanding and social navigation lay the groundwork for robots capable of acting effectively, safely, and adaptively, but above all in a way that is accepted and appreciated by people in real-world contexts.
L’integrazione dei sistemi robotici negli ambienti condivisi con le persone richiede che i robot non siano soltanto sicuri e funzionali, ma anche socialmente accettabili. L’accettazione sociale viene definita come il grado in cui gli utenti si sentono al sicuro, compresi e a proprio agio nell’interazione con un robot, e dipende da fattori quali fiducia, trasparenza, rispetto delle norme sociali e adattabilità. La tesi affronta tale tema proponendo inizialmente un modello di riferimento fondato su tre componenti: definizioni, requisiti e architettura. I requisiti comprendono aspetti funzionali e non funzionali, quali usabilità, utilità percepita, sicurezza, privacy e allineamento culturale; dal punto di vista architetturale, un robot socialmente accettabile deve integrare percezione multimodale, consapevolezza del contesto che lo circonda, sicurezza e controllo adattivo. All’interno di questo modello, il lavoro di ricerca di questo elaborato si concentra su due aspetti principali: comprensione dell’utente e navigazione sociale. La comprensione dell’utente riguarda la capacità del robot di rilevare e interpretare stati cognitivi ed emotivi, per personalizzare l’interazione e incrementare comfort e fiducia con l'utente. A tale scopo, sono stati sviluppati metodi di classificazione basati su sensori multimodali e analisi di biosignali (EEG, ECG, GSR, espressioni facciali), finalizzati a stimare carico mentale e livelli di stress in scenari di collaborazione uomo-robot. Gli studi hanno evidenziato la validità di indici fisiologici come predittori dello stato cognitivo ed emotivo, e hanno portato alla realizzazione di un dataset multimodale pubblico dedicato alla robotica collaborativa. Questi strumenti consentono al robot di adattare dinamicamente parametri operativi migliorando la qualità dell’interazione. La navigazione sociale, fondamentale per i robot mobili in ambienti dinamici e popolati, è stata affrontata attraverso lo sviluppo di un framework di controllo capace di integrare vincoli spaziali e temporali assicurando che il robot abbia comportamenti socialmente accettabili. L’approccio combina logiche temporali dei segnali (STL) con funzioni di barriera di controllo (CBFs), garantendo il rispetto delle specifiche e la sicurezza dell’interazione anche in condizioni di incertezze e presenza di ostacoli dinamici. Il framework è stato esteso con vincoli addizionali per il mantenimento del campo visivo dell’utente e con formulazioni robuste che integrano tecniche di filtraggio, assicurando affidabilità anche in scenari reali complessi. Parallelamente, è stata proposta un’architettura cognitiva che integra modelli linguistici di grandi dimensioni (LLMs) per interpretare input eterogenei (stato emotivo, intenzioni, contesto sociale) e tradurli in parametri di navigazione, arrivando così al concetto di navigazione sociale intelligente. Questa soluzione abilita il robot ad adattare la propria strategia di movimento in modo dinamico e contestuale. Inoltre, è stata sviluppata un’architettura leggera per compiti di guida sociale, basata su riconoscimento facciale, rilevamento gestuale e tracciamento scheletrico, in grado di garantire un comportamento trasparente e reattivo in ambienti pubblici. Nel complesso, la tesi dimostra che l’accettazione sociale non è una proprietà statica del robot, ma un fenomeno emergente dall’interazione con l’utente. Le soluzioni proposte per la comprensione dell’utente e la navigazione sociale pongono le basi per robot capaci di agire in modo efficace, sicuro e adattivo, ma soprattutto accettato e apprezzato dalle persone nei contesti reali di utilizzo.
Accettazione sociale nella robotica: dalla comprensione dell’utente basata su biosignali alla navigazione sociale intelligente
RUO, ANDREA
2026
Abstract
The integration of robotic systems into environments shared with people requires that robots are not only safe and functional, but also socially acceptable. Social acceptance is defined as the extent to which users feel safe, understood, and comfortable when interacting with a robot, and depends on factors such as trust, transparency, compliance with social norms, and adaptability. This thesis addresses the topic by first proposing a reference model based on three components: definitions, requirements, and architecture. The requirements include both functional and non-functional aspects, such as usability, perceived usefulness, safety, privacy, and cultural alignment. From an architectural perspective, a socially acceptable robot must integrate multimodal perception, awareness of the surrounding context, safety, and adaptive control. Within this model, the research presented in this work focuses on two main aspects: user understanding and social navigation. User understanding concerns the ability of the robot to detect and interpret cognitive and emotional states, in order to personalize interaction and enhance user comfort and trust. To this end, classification methods based on multimodal sensors and biosignal analysis (EEG, ECG, GSR, facial expressions) were developed to estimate mental workload and stress levels in human-robot collaboration scenarios. The studies confirmed the validity of physiological indices as predictors of cognitive and emotional states and led to the creation of a public multimodal dataset dedicated to collaborative robotics. These tools allow the robot to dynamically adapt operational parameters, thereby improving the quality of interaction. Social navigation, which is crucial for mobile robots operating in dynamic and populated environments, was addressed through the development of a control framework capable of integrating spatial and temporal constraints to ensure socially acceptable behavior. The approach combines signal temporal logic (STL) with control barrier functions (CBFs), guaranteeing compliance with specifications and safety of interaction even under uncertainty and in the presence of dynamic obstacles. The framework was extended with additional constraints to maintain the user within the robot’s field of view and with robust formulations that incorporate filtering techniques, ensuring reliability even in complex real-world scenarios. In parallel, a cognitive architecture was proposed that integrates large language models (LLMs) to interpret heterogeneous inputs (emotional state, intentions, social context) and translate them into navigation parameters. This solution enables the robot to dynamically and contextually adapt its movement strategy, leading to the concept of intelligent social navigation. In addition, a lightweight architecture was developed for social guidance tasks, combining facial recognition, gesture detection, and skeleton tracking, in order to guarantee transparent and reactive behavior in public environments. Overall, the thesis demonstrates that social acceptance is not a static property of the robot, but an emergent phenomenon resulting from interaction with the user. The proposed solutions for user understanding and social navigation lay the groundwork for robots capable of acting effectively, safely, and adaptively, but above all in a way that is accepted and appreciated by people in real-world contexts.| File | Dimensione | Formato | |
|---|---|---|---|
|
Ruo.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
67.94 MB
Formato
Adobe PDF
|
67.94 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/20.500.14242/362898
URN:NBN:IT:UNIMORE-362898