Understanding and modeling human behavior are major challenges in virtual reality (VR) research, enabling technological advances (e.g., foveal rendering, which concentrates computing power in areas of visual attention, or the creation of realistic virtual agents) as well as scientific advances in the humanities and social sciences (e.g., the study of group behavior).
Literature studies seeking to understand and model behavior have mainly focused on visual attention [Sitzmann et al. 2018, Goudé et al. 2023] and movement [Bourgaize et al. 2024, Patotskaya et al. 2024]. However, cross-sectional studies of gaze and movement remain largely unexplored and tend to focus on specific cases [Raimbaud et al. 2023] rather than on a series of cases that would allow the construction of a predictive model. Furthermore, these studies have mainly considered visual stimuli only (salient elements for gaze, obstacles for movement) and, in a few recent cases, spatialized auditory stimuli [Bernal-Berdun et al. 2023, Bernal-Berdun et al. 2024].

The objective of this thesis project is to study human behavior when immersed in a multisensory virtual environment that includes visual, auditory, thermal, and olfactory stimuli. The planned steps are: i) an experimentation phase to collect behavioral data (motion capture, eye tracking, physiological signals) during VR scenarios; ii) an analysis and behavior modeling phase using artificial intelligence tools; iii) the use of these models to predict user behavior in VR and develop virtual agents with reactive behaviors.