Integrating internal performance measures into the decision making process of autonomous agents

Abstract

Integrating performance measures into the process of decision making of an autonomous agent is a common method in artificial intelligence. Reinforcement learning is one possible application that can be realized with this methodology. Recent findings in artificial intelligence showed the importance of the body and the tight connection to the decision making processes. This article introduces a model for integrating internal performance measures into such a decision making process. The introduced theory is based on psychoanalytical concepts that are technically specified and realized in a virtual robot, situated within a simulated environment. Furthermore, it is shown how the concept can be applied to already existing cognitive architectures and their implementations, such as the BDI (Belief-Desire-Intention) architecture. The results show the dynamics in decision making that become possible with the newly applied performance measures.