Learning of Object-Action Complexes

PP Restricted to other programme participants (including the Commission Services) RE Restricted to a group specified by the consortium (including the Commission Services) CO Confidential, only for members of the consortium (including the Commission Services) This deliverable is the continuation of the last year deliverables D2.1 and D2.2 in which we described algorithms and software that implements basic sensorimotor processes for learning of object-action complexes. In this deliverable we show the videos of the current state of implementation of these processes. More specifically, we show the following videos: • oculomotor control implementing foveation, • visual attention on a humanoid robot, • learning appearance-based object representations for recognition, and • human motion capture with the active head. In the report we briefly describe the videos and the underlying algorithms. While some of the algorithms have been developed already in the first year of the project, we have now implemented all of them on a humanoid robot ARMAR and on the copies of the ARMAR’s humanoid head