We view cognition as the ability to foresee the future course of action and plan our behavior consequently1. Cognition develops at the interface between brain processes on one side and forceful actions at the other: that is, bodily actions are necessary for the development of cognition in humans and, in our view, also in the intelligent machines of the future. In particular, manipulation represents one of the strongest means to access knowledge about the world in the form of syntax and semantics which can be also applied to language2. We work on both sides of this research by building bodies for cognition.
The Bodyware
The bodyware is the "physical infrastructure" behind the cognitive abilities. Neuroscientists tell us that the brain depend on the body, the robot's mindware ought to depend on what shape and technologies we can deploy on the robot's body.
 |
 |
 |
 |
 |
 |
The Mindware
The mindware is the "juice" of the cognitive abilities (or will be). Based on the "reverse engineering" of the brain, on mathematical modeling, and on full-fledged implementations on the iCub, we aim at providing intelligence to our machines.
 |
|
|
|
|
|
|
Perception |
Interaction |
Bodyware: the iCub
The iCub is the humanoid robot developed as part of the EU project RobotCub and subsequently adopted by the Cognitive Humanoids Laboratory at IIT. It has 53 motors that move the head, arm & hands, waist, and legs. It can see and hear, it has the sense of proprioception (body configuration) and movement (using accelerometers and gyroscopes). We are working to improve on this in order to give the iCub the sense of touch and to grade how much force it exerts on the environment. Simultaneously, we are already planning the next generation of humanoid robots: they will be made of soft actuators, new stronger and lighter joints with power efficiency closer to that of biological systems like ourselves.
Mindware: the technology of cognitive systems
Our real interest is of course in the eventual realization of intelligent machines. We research on the basic technology of cognitive systems and on its translation into a full working system. Our modus operandi in this case starts from biology, that is by "reverse engineering" the brain, continues through mathematical modeling of specific brain functions that finally find an implementation on the humanoid robot. Examples of the cognitive skills that we study are attention, reaching and manipulation. Although broad in scope this research has a clear and measurable target also of industrial interest. In particular, humanoid robots can be imagined as helpers in manufacturing, office or home environments. In this respect, for example, we address safety in human-machine interaction (at the cognitive, control, and hardware level). We also participate to projects that will add language skills to the ICub and we are developing microelectronics for sensing, actuation and processing.
1. von Hofsten, C., An action perspective on motor development. Trends in cognitive sciences, 2004. 8(6): p. 266-272.
2. Fadiga, L., et al., Speech listening specifically modulates the excitability of tongue muscles: a TMS study. European Journal of Neuroscience, 2002. 15(2): p. 399-402.
