Computation mechanism for situated sentient robot

Emotions shape how we remember our world, how we perceive it, and which decisions we take. This study proposed a novel three-component computation mechanism that enables a robot to reason about emotions. The mechanism contains the following parts: (i) information acquisition, (ii) formal knowledge description in a form of ontology, and (iii) Bayesian Network (BN). An entropy reduction method from information theory is used to design an effective Human-Robot Interaction (HRI) and to gain a deeper understanding about proposed reasoning mechanism. The method also revealed the most influential BN variables that efficiently resolve the reasoning ambiguities. The modified OCC model of emotions in BN is implemented to ensure adaptation of the system to multiple sources of uncertainty. Variables in a hand crafted BN are linked together, where each link is quantified by spreading influences of a different strength that parent nodes have on child nodes. The paper introduces also the system architecture for realization of the physical robot setup.