A cognitive approach for a humanoid painter

In this work we propose a system which exploits a cognitive architecture in order to implement a creative painting mechanism in a humanoid robot such as the Aldebaran Nao. The system is capable of autonomously assessing its own work during the portrait creation process, which is simultaneously influenced by the perceptions of the robot. In particular it takes into account the perception of the visual appearance of the portrait which is being realized. The system output is ready to be converted in a set of Nao arm gestures. This would let the Nao robot capable of drawing a portrait by waving some sort of virtual pen in the air. The Nao gestures will be captured through an infrared acquisition system and interpreted as being movements of a brush on a canvas. The results will be continuously evaluated and compared with the internal model of portrait in order to decide how to complete the painting. This approach will realize an autonomous dynamic improvement process, and not a mere planning painting procedure.