Integrating Agent Planning and Emotions

Abstract

Agent planning and emotion simulation are two aspects of agent based research that have largely been kept separate. This is despite plenty of literature showing the significant impact that emotions have on human decision making. In this thesis we determine the feasibility and benefits of generic integration between basic emotion systems and planners. We believe that by enabling interaction between emotions and automated planners we can more accurately mimic the human decision making process, allowing for more realistic solutions to a potentially greater set of problems. From our research we have derived a model to facilitate the integration of existing emotion and planning systems. The model specifies how, several basic planning components such as actions and goals, can be used to allow emotions to influence, and be influenced by, a planner. Lastly, we implemented the model as a prototype which enhances a planner with emotions from an emotion system. Our results show that a basic scenario with emotions such as anger and fear can already enhance the planning significantly. It also enabled the planner to resolve a social scenario using purely emotional reasoning. Lastly, we discuss the cases in which we believe the model to be applicable.