From robot exploration to the statistics of gaze, a research review and future directions in cognitive vision Abstract: In this seminar I will first review my past research activity on cognitive robotics and progressively shift towards my current research interest on the computational modeling of visual attention and finally I will introduce future directions that I plan to pursue.I focused my research mainly on feature extraction and analysis of sensorymotor data from a human tutor with the aim of modelling interesting and paradigmatic behaviours in scene viewing. To this purpose we developed a custom wearable device able to track simultaneously the human gaze and the full head pose1. Collected data are supposed to be informative about the way humans deploy attention on the surrounding environment, especially under a task commitment. These efforts were framed in the long-range perspective of the construction of a biologically inspired model of attention, according to which a robot or an artificial vision system would be able to focus on (task) relevant portions of the perceived scene. In so doing the allocation of visual processing ressources is optimised and higher level cognitive functions such as object/shape detection and recognition, decision making, navigation, action/planning are speeded up.

This document describes the status of progress for the research on learning skills for functioning processes and task execution performed by the NIFTi Consortium. In particular, according to the Description of Work (DOW), research is focused on the development of novel methods and paradigms for skill learning based on affordances and action-reaction observation. Planned work, as per the DOW, is introduced and the actual work is discussed, highlighting the relevant achievements, how these contribute to the current state of the art and to the aims of the project. 1 Methods and paradigms for skill learning Gianni, Papadakis, Pirri, Pizzoli