I am doing PhD research in composition of actions and language for iTalk project, which aims to develop artificial embodied agents able to acquire complex behavioural, cognitive, and linguistic skills through individual and social learning.

For the first part of my research I am using a special type of neural network developed by Jun Tani that has different types of neurons that allow the network to emerge a hierarchical structure distinguishing motor primitives from entire sequences of actions. The network has recurrent connections and time integration through time and already mentioned multiple timescales. At this stage, the network is trained (using backpropagation through time) simple sequences of actions, such as lifting a box 3x up/down and moving it 3x left right, which is then tested on the simulated robot and also on our real humanoid iCub platform. Once these experiments are over, I will start adding a language component to the network to investigate how language and actions co-develop, how one affects the other or bootstraps its development.