Thursday Feb 24, 1:30, WeH 4601 Dealing with Huge Amounts of Data in an Object Reaching Task Filippo Neri For human beings, the easiest way to recognize objects and to realize where they are is just watching the world. As we realized how powerful and important is the sight capability in operating in our world, we thought that also an artificial agent, that was able to exploit such a capability, may efficiently accomplish a wide variety of tasks. The main problem, to enable an artificial agent to 'see', is dealing with the huge amount of data coming from its sensors. In human beings, for instance, the signals coming from our eyes have different importance/meaning depending on what we are currently doing. If you are hungry and you are stepping to your fridge, generally you don't notice the pictures hanging on the wall. So, roughly speaking, your brain considers as noise the signals relative to anything else but a fridge. In my talk, I'll describe a focalization of attention methodology for the analysis of images that may be used, by an artifial agent, in tasks like object reaching or landmark recognition. So far, only experiments in simulation have been performed. I'll also discuss some of them.