Game Over: You have been beaten by a GRUE

We describe GRUE, an architecture for game agents. GRUE facilitates the development of flexible agents capable of balancing competing goals and responding appropriately to their environment. We briefly describe GRUE’s representation of the agent’s environment and sketch the algorithm used to select between competing goals. We present some preliminary results from an evaluation of GRUE in a Tileworld environment which suggest that the features of GRUE are useful in dynamic environments. We conclude with a brief description of future work.