Object Recognition through Manipulation


We aim to implement human inspired haptic exploration strategies on the iCub. To do so, we collected a unique dataset with an unprecedented variety of sensors monitoring exploratory actions of blind-folded subjects performing an object recognition task. We will study the relationship between the information acquired during the exploratory procedures and the task performance to inform an artificial model on how to perform informative exploratory procedures with the iCub.

The recorded dataset (available soon upon request) includes the hand kinematics and contact force trace measured with the Cyber Glove prototype (from Bielefeld University), the hand and arm kinematics recorded using optical tracking (Vicon), and the muscular activity from EMG of the arm (COMETA bi-polar EMG) and lower arm (OTB eletronica high-density EMG).

Experiments with healthy subjects, sensor fusion, multimodal, muscle synergy decomposition, optical tracking, EMG, tactile glove.