Robot´s emotion generation model for transition and diversity using energy, entropy, and homeostasis concepts

Abstract-This study is for describing a robot´s emotion generation and transition by introducing concepts of energy, entropy, and homeostasis. Robot has been deeply involved to the human society, and required social roles to interact with people personally, socially, and universally. So a lot of researchers have been studied how a robot can generate their own emotions. Among several approaches to explain the emotional structure, the dimensional approach based on factors of pleasure and arousal is used due to the merit of the interpolation between emotions in an emotional space. Specifically, Circumplex model is used which has also two factors: pleasure and arousal, and this model indicates how emotions are distributed in the two-dimensional plane. Two kinds of energy states: mental energy and physical energy are introduced by the definition of psychological energy from psychodynamics. Then these two energy states are matched to pleasure and arousal respectively. First, the mental energy is updated by the result of Prospect theory which measures the value of gain and loss as pleasure factor. And the physical energy is updated by the result of the hedonic scaling which measures the levels of arousal from pleasure, and by the intensity of stimuli. Then each energy state is fed by entropy and homeostasis. The feedback loop by entropy is to satisfy the 2nd law of thermodynamics, and the feedback loop by homeostasis is to satisfy the property of a living system. The energy states generated by stimuli and fed by entropy and homeostasis take a position in the plane of Circumplex model. And distances between the current position and other emotions are computed to get a level of each emotion, proportional to the inverse of the distance. Finally, for the experimental result, only Ekman´s six basic emotions were expressed by the face simulator, FRESi to which LDAEM is applied.