| CREATE: Community Robotics, Education and Technology Empowerment
Carnegie Mellon University
[Contact Information | Introduction | Projects | People ]
|Lab:|| CREATE: Community Robotics, Education and Technology Empowerment
Newell-Simon Hall A504 (MSL)
|Contact:|| Illah R. Nourbakhsh
Carnegie Mellon University
Pittsburgh, Pennsylvania 15213
We feel it is possible to find ways to apply abstraction and hierarchy to robots by understanding how humans reason with these concepts. To achieve this, we are developing a Representation Designer.
The goal is to design a generic graphical tool that will allow human users to define a representation of the world and test it with planning and diagnosis problems. This representation will incorporate objects, actions, goals, time and the relationships between these entities.
Our idea is to make this representation non-rigid and incorporate hierarchy. For example:
If you tell the robot that you're going to a lab meeting at 4:00, it calls you at 3:00 at home to remind you to go to CMU and brings your lab notebook to the meeting for you.Click here to see a screen shot from the current working interface!
For forty years, computer scientists have had the field of Computer Architecture to provide a framework in which to describe their work. A new system or piece of hardware can be described in terms of well defined terms and relationships. In order to describe the internal workings of a robot, however, people have to resort to extensive sketches, ad hoc descriptions and artistic hand gestures.
We are grabbing robotics by its roots in Computer Science, Control Systems, Electrical Engineering and other fields, and constructing a formal descriptive and analytic methodology. We are applying the principles from already established formalisms including Real-Time Multimedia Systems, Feedback Control Theory, Petri Net Theory, Computer Architecture, and apply them to the subfields of robotics, ranging from Intelligent Agents down through Actuator Control.
Learning Hierarchies of Abstraction
In order to allow robot learners to solve more complex problems with a minimum of human intervention, we are developing a system that automatically generates a hierarchy of behaviors, useful subtasks, and abstractions from real-world experience. The system uses a model of attention to choose subgoals and guide exploration such that creating abstractions is easier.
Sal Desiano is a Robotics PhD student at the Robotics Institute and is concentrating on the holy grail of Robot Architecture.
Carlos Guestrin is a visitor at the Robotics Institute and is focusing on the Representation Designer project and Educational Robotics. Carlos is also involved in Visual Position Estimation research.
Illah R. Nourbakhsh is Assistant Professor of Robotics at the Robotics Institute. He runs the Mobile Robot Programming Lab and is also a member of the Automated Highway System project.
Iwan Ulrich is a Robotics PhD student at the Robotics Institute and leads the Bookstore Project.
[ The Robotics Institute | Carnegie Mellon University ]
|Last modified September 29, 1997 Illah R. Nourbakhsh ( email@example.com)|