The modeling of agents involving emotions and affective states constitutes a relevant discussion topic in the research concerning multi-agent simulations, especially because of how the introduction of affective parameters inside the modeling process could effectively make the produced simulations more realistic. In this research area, though, the modality in which parameters regulating the affective state of agents are introduced into models, so that the agents’ behaviour and actions are influenced by them, is always based on emotional models found in literature, or on physics theories and models usually involved for the modeling of pedestrians and crowds. The approach this work presents, then, aims at tackling this problem from the point of view of data, thus wanting to get to affective agent modeling starting from data coming from real people, acquired through ad-hoc experiments with the precise goal of observing reactions and behaviour to be later translated inside a model. In particular, the focus of this work falls on the research on pedestrians and walkability, observing different types of interactions involving pedestrians through four different experiment through which gather data able to describe the participants’ interactions to then implement them in the modeling step. The proposed experiments are executed in-vivo, in-vitro and online, observing pedestrian interactions with vehicles, moving obstacles and other pedestrians, gathering data regarding these interactions through physiological data and questionnaires made for profiling purposes and in order to have more information regarding the subjects’ behaviour and reactions. The gathered data is then used for modeling, firstly from the point of view of cellular automata and then passing on to the multi-agent systems perspective, showing how the information obtained from the data is introduced inside the models to be parametrized in affective parameters that, depending on the assigned values, could influence in a certain way the behaviour of the agents. After that, some simulation instances derived from the models are presented, as to observe how the affective parameters that were introduced in the models actively influence the behaviour of agents acting and moving in certain situations.

La modellazione di agenti che tiene conto di emozioni e stati affettivi costituisce un argomento di discussione piuttosto importante nell’ambito della simulazione ad agenti, soprattutto per via di come introdurre parametri affettivi nella modellazione possa contribuire a rendere le simulazioni più realistiche. In questo ambito di ricerca, però, il modo di introdurre nei modelli parametri in grado di regolare lo stato affettivo degli agenti per da influenzarne azioni e comportamenti è spesso basato sui modelli emozionali che si trovano in letteratura, oppure sulle teorie e i modelli fisici che vengono solitamente utilizzati per la modellazione di pedoni e folle. L’approccio presentato in questo lavoro, quindi, mira ad approcciare il problema dal punto di vista dei dati, puntando ad arrivare alla modellazione di agenti affettivi partendo da dati provenienti da persone reali ed acquisiti tramite esperimenti creati ad hoc con il preciso obiettivo di studiare reazioni e comportamenti da poter poi tradurre in modellazione. In particolare, in questo lavoro il problema viene affrontato concentrandosi in particolare sull’ambito pedonale, osservando diversi tipi di interazione coinvolgenti pedoni tramite quattro diversi esperimenti atti a raccogliere dati in grado di descrivere le interazioni operate dai soggetti per poi inserirle in un contesto di modellazione. Gli esperimenti vengono effettuati in vivo, in vitro e online, osservando le interazioni di pedoni con veicoli, ostacoli in movimento ed altri pedoni, raccogliendo dati riguardo queste diverse interazioni tramite dati fisiologici e questionari atti a profilare i partecipanti e a fornire maggiori informazioni riguardo al comportamento e alle reazioni da loro dimostrate. I dati raccolti vengono quindi utilizzati per la modellazione, prima in ambito di automi cellulari e poi, successivamente, nell’ambito dei sistemi multi-agente, mostrando come le informazioni ricavate dai dati vengano integrate all’interno dei modelli al fine di includere parametri affettivi che, in base ai valori assegnati, influenzino in un certo modo il comportamento degli agenti. Vengono poi proposte alcune simulazioni derivanti dai modelli, ai fini di osservare come i parametri affettivi introdotti influenzino il comportamento degli agenti in azione in determinate situazioni.

(2023). From Real Affective States towards Affective Agents Modeling. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2023).

From Real Affective States towards Affective Agents Modeling

GILTRI, MARTA
2023

Abstract

The modeling of agents involving emotions and affective states constitutes a relevant discussion topic in the research concerning multi-agent simulations, especially because of how the introduction of affective parameters inside the modeling process could effectively make the produced simulations more realistic. In this research area, though, the modality in which parameters regulating the affective state of agents are introduced into models, so that the agents’ behaviour and actions are influenced by them, is always based on emotional models found in literature, or on physics theories and models usually involved for the modeling of pedestrians and crowds. The approach this work presents, then, aims at tackling this problem from the point of view of data, thus wanting to get to affective agent modeling starting from data coming from real people, acquired through ad-hoc experiments with the precise goal of observing reactions and behaviour to be later translated inside a model. In particular, the focus of this work falls on the research on pedestrians and walkability, observing different types of interactions involving pedestrians through four different experiment through which gather data able to describe the participants’ interactions to then implement them in the modeling step. The proposed experiments are executed in-vivo, in-vitro and online, observing pedestrian interactions with vehicles, moving obstacles and other pedestrians, gathering data regarding these interactions through physiological data and questionnaires made for profiling purposes and in order to have more information regarding the subjects’ behaviour and reactions. The gathered data is then used for modeling, firstly from the point of view of cellular automata and then passing on to the multi-agent systems perspective, showing how the information obtained from the data is introduced inside the models to be parametrized in affective parameters that, depending on the assigned values, could influence in a certain way the behaviour of the agents. After that, some simulation instances derived from the models are presented, as to observe how the affective parameters that were introduced in the models actively influence the behaviour of agents acting and moving in certain situations.
ANTONIOTTI, MARCO
GASPARINI, FRANCESCA
BANDINI, STEFANIA
stati affettivi; esperimenti; pedoni; agenti; automi cellulari
affective states; experiments; pedestrians; agents; cellular automata
INF/01 - INFORMATICA
English
17-feb-2023
INFORMATICA
35
2021/2022
open
(2023). From Real Affective States towards Affective Agents Modeling. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2023).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_795267.pdf

accesso aperto

Descrizione: Tesi Finale Giltri Marta PhD - Minor Revision
Tipologia di allegato: Doctoral thesis
Dimensione 19.61 MB
Formato Adobe PDF
19.61 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/404517
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact