Designing a Growing Functional Modules “Artificial Brain”

Abstract

The present paper illustrates the design process for the Growing Functional Modules (GFM) learning based controller. GFM controllers are elaborated interconnecting four kinds of components: Global Goals, Acting Modules, Sensations and Sensing Modules. Global Goals trigger intrinsic motivations, Acting and Sensing Modules develop specific functionalities and Sensations provide the controlled system's feedback. GFM controllers learn to satisfy some predefined goals while interacting with the environment and thus should be considered as artificial brains. An example of the design process of a simple controller is provided herein to explain the inherent methodology, to exhibit the components' interconnections and to demonstrate the control process.