First Investigations into Artificial Emotions in Cognitive Robotics

摘要

In nature, the combination of processes of emotion and cognition has a deep impact on type and quality of reaction to environmental stimuli. In this work, we want to test the feasibility of artificial hormones in artificial neural networks. We take a minimal evolving neural network and look into the implications and opportunities of extending this model of communicating nodes, with one virtual hormone gland. To explore the differences in behavior, that we expect to develop with this modification, we modify an already well established model, the Braitenberg Vehicle. These vehicles were faced with a simple energy gathering task. The behavior, efficiency and fitness of these vehicles in identical environment, with the artificial hormone active and inactive, is examined. It shows, that the implementation of artificial emotion leads to an increase in efficiency of the evolved solution.