Current page: Members->List of Members->Fiora Pirri
 

Fiora Pirri
University of Rome "La Sapienza"
Membership Number: 35
Address: Dipartimento di Informatica e Sistemistica, Via Salaria 113, 00198 Roma, Italy
Email: pirri@dis.uniroma1.it
Phone: +39 06 4991 8338
Fax: +39 06 5830 0849
URL: http://www.dis.uniroma1.it/~pirri/

Biographical Sketch
Fiora Pirri, Associate Professor

University of Rome "La Sapienza" - ALCOR: Autonomous Agent Laboratory for Cognitive Robotics
The Autonomous agents Laboratory for Cognitive Robotics begun the research activity in Cognitive Robotics around 1998, coming from the areas of Knowledge Representation and Reasoning in Artificial Intelligence. Since 1998, our research has collected several disciplines from Artificial Intelligence, in particular those involving theories of action and perception, planning, reasoning and automated deduction, decision making and learning. In the last two years we have understood that, for the implementation of a fully cognitive (robotic) agent, the cognitive level had to be involved in the visual recognition process. Starting from this personal opinion, we have studied a way to translate the information caming from the vision system into a symbolic representation, in order to allows its interpretation using a specific reasoning process.

We work with mobile manipulator robots (MoMaBot) endowed with a rich sensory system, in particular vision. We have realized a control system built according to a cognitive multi layered architecture integrating a cognitive, reactive and structural control.

A cognitive architecture is a formal, structural and computational model specifying the knowledge, the behavior and the functionalities which are necessary to a dynamical system for autonomously interacting with the environment, and with other agents, when required. A cognitive architecture is essential for the realization of a robot able to perform multiple tasks in the real world, and capable to handle several unforeseen problems, that he encounters while trying to reach his goals.

We have implemented this architecture on several tiny robots. The last one is ArmHand1, spelled Armandone, but we are now constructing a new cognitive robot named ArmHandX, which is enrolled to the Robocup Resque. Armandone crucial abilities are sensing (via sonars, cameras, and encoders), processing the acquired information for building a model of the environment, and inferring properties about the domain. Here the inference uses the robot’s logical attitudes. Armandone’s skills, in fact, allow him to find a strategy, challenged by imprecise sensors and actuators, for traveling around and for successfully completing suitable tasks.

Armandone has participated to the AAAI-2002 Robot-Exhibition and Competition, in Edmonton Canada in July-August 2002. The task at the exhibition is to explore a completely unknown maze, picking up objects in the hallways, and find the exit. Armandone is, in fact, an autonomous mobile robot roaming non engineered indoor mazes. The real novelty about Armandone consists in his initial knowledge about the world. In fact, since Armandone is prepared to explore a completely unknown environment, his prior knowledge is not concerned with the map of the environment, objects, positions, etc. Armandone prior knowledge is concerned exclusively with physical, structural, topological, and common sense laws ruling his environment.

All the information about the world are captured using sonar sensors and a couple of stereo cameras, and are fused into a representation that is used by the high level reasoning system to take a decision.

Finally, our long experience in cognitive robotics, the backgroud skills of the group and the strong working relationship that we have with other european group is our immediate contribution to the EcVision Network.


Site generated on Friday, 06 January 2006