From Sensable to Sensible Spaces: Enhancing the Sensibility of a Home Office using Stress-Aware Deep Reinforcement Learning in Virtual Environments

Emanuel Gollob, Maria Kyrou, Panagiotis C. Petrantonakis, Ioannis Kompatsiaris

Abstract: Being confronted with a dulling perception and challenges to be creative when working at home, I imagined my home office to sense my state of arousal and search together with me for inspirational moments by infusing its familiar appearance with distortion. In two prototypical human-home office interaction systems, we use Virtual Reality (VR), Deep Reinforcement Learning (DRL) and Galvanic Skin Response (GSR) to enhance a home office’s sensibility towards its user’s level of arousal as well as to enlarge its textural action space. Although physiological feedback in machine learning faces low learning rates, the resulting interaction offers a fresh perspective on our human-home office relation.