Multilevel Darwinist Brain: Context Nodes in a Network Memory Inspired Long Term Memory

The Multilevel Darwinist Brain (MDB) is a cognitive architecture aimed at providing autonomous and self-motivated life-long learning capabilities for robots. This paper deals with a new structure and implementation for the long term memory (LTM) in MDB based on Fuster’s concept of Network memory and on the introduction of a new type of node or cognit called Context Node (Cnode). The idea of Network memory as proposed here, provides a path to hierarchically and progressively relate LTM knowledge elements, allowing for a developmental approach to learning that permits very efficient experience based responses from the robot. We include a simple, albeit quite illustrative, example of the application of these ideas using a real Baxter robot.