To endow virtual agents with more human-like behaviour, the current project proposes a new model for decision making based on emotions and trust. The project focuses first on the conceptual design of the model, then on the implementation of this model. To try to reproduce human decision making, the model has, mostly, been based on psychologically grounded findings. This model is based on the BDI structure, and has been extended with submodels for emotions and trust. To further prove the implementability of the model, it was implemented first - at a conceptual level - in LeadsTo, and thereafter in the RoboCup soccer environment. The implementation in LeadsTo shows exactly how the connections and information flows of the model work. The results demonstrate that agents using emotions to bias their decisions act differently than the agents that use rational thinking. To investigate these results further, and in a more dynamic environment, the model was then successfully implemented in the RoboCup soccer environment, 2D simulation. Some of the results were significant differences in actions between players with different goals, and differences in some of the performed actions between a positive versus a negative emotional team.
|Publication status||Published - 2010|