Architecture and structured representations for an autonomous agent

The current work illustrates our approach of artificial autonomy integrating reactive and goal oriented capacities. A three layered architecture (physical, behavioural and cognitive) is described as a solution to the action selection problem in artificial agents modelling. Particular attention is given to the cognitive layer including a two stages structured knowledge representation-the sensorimotor history and the topological interaction-and three cognitive processes-interpretation, learning and motivation-which are responsible for structuring and exploiting the representations, in order to increase the agent´s autonomy. This model has been successfully applied to our actual mobile robot for navigation tasks.