Bayesian networks for representing affective agent based on situation calculus

Abstract

Intelligent robots or autonomous agents that could express human-like emotions are more trustworthy when humans communicate with them. Agents choose appropriate actions not only depending on current goals and the information perceived from the environment, but also on their mental state. We take formal accounts of how to represent affective states of agents when reasoning about actions and their effects. We discuss the computational description by mapping our model onto the Bayesian networks. In this paper, we propose a logical approach that clearly illustrates the relations among actions, situations, sensing, and affective states in the situation calculus.