A Reactive Competitive Emotion Selection System

We present a reactive emotion selection system designed to be used in a robot that needs to respond autonomously to relevant events. A variety of emotion selection models based on “cognitive appraisal” theories exist, but the complexity of the concepts used by most of these models limits their use in robotics. Robots have physical constrains that condition their understanding of the world and limit their capacity to built the complex concepts needed for such models. The system presented in this paper was conceived to respond to “disturbances” detected in the environment through a stream of images, and use this low-level information to update emotion intensities. They are increased when specific patterns, based on Tomkins’ affect theory, are detected or reduced when it is not. This system could also be used as part of (or as first step in the incremental design of) a more cognitively complex emotional system for autonomous robots.