What will students learn in your course?
I teach several artifical intelligence methods within the programme's first year. In essence, in my course students learn how to make machines solve complex problems. In order to give machines the capability to solve such problems, we first need to describe them with a useful representation, and then we need the ability to figure out how to solve the important parts of the problem using that representation.
One example is playing chess. This is something that a computer, or even your phone or watch, can do quite well — even though it has to choose among more possible moves than there are atoms in the universe! The trick is to use an appropriate problem representation and use it to make informeddecisions.
Another example is making a robot set a table: placing a knife, a fork, and a dish. For most humans, this is much easier than beating the world champion in chess. However we still have trouble making a robot that can do this as well as we can. Of course, we can program the motions of a robot to set a specific table under specific conditions; but giving the robot the general capability to "set tables" (including tables it has never seen before, in places it has never been before) is still a great scientific challenge. This is because acting in the real world involves solving many problems concurrently: the robot must the items to the table without colliding with objects or people, it must observe the table and understand where to place the objects that are missing, it has to plan when and how to perform each task (first grasp, then move, then put down), etc. To address these problems, we need AI methods such as the ones that I teach in this course.
These AI methods are of course useful in many other domains too, such as robots in a work environment, and computer games often use AI techniques to plan the actions of characters.
Why have you chosen to carry out your research in Örebro?
I came to Örebro in 2008 from the Italian National Research Council. I chose the Center for Applied Autonomous Sensor Systems (AASS) at Örebro University because it is one of the few places in Europe, or even the world, where people who know about robotics and AI also know a bit about each others' disciplines. This is very stimulating for integrating the two domains. Thanks to its multi-disciplinary and highly competent team of researchers, AASS offers an unique opportunity for developing intelligent robots that operate in real applications.
What is your research about?
My research interests lie at the intersection of Artificial Intelligence and Robotics. I focus specifically on constraint reasoning, planning and scheduling, and meta-CSP techniques for hybrid reasoning.
Much of my work nowadays focuses on developing constraint reasoning algorithms for robot planning and for context recognition from sensor traces. In the past five years, I have been applying these techniques in two broad application areas: service robots/sensor systems for use in domestic environments; and decision support tools for industrial scenarios with large autonomous vehicles. I am fascinated by the problem of using AI for robots, and I find that some of the most compelling research questions in AI originate from the use of model-based approaches to robot control.