David Touretzky

David Touretzky

Research Professor

Office: 9013

Email: dst@cs.cmu.edu

My current research focuses on robotics and computing education. I have previously done work in computational neuroscience, connectionist modeling, and knowledge representation.

Tekkotsu Framework for Robotics Education

When robots incorporate vision systems and high degree-of-freedom manipulators, simple reactive programming strategies no longer suffice. How can state of the art robot programming be made accessible to beginners? Tekkotsu ("framework" in Japanese) is an open source software framework used to introduce computer science undergraduates to essential robotics topics such as computer vision, landmark-based navigation, path planning, and manipulation. I am interested in how robot programming systems can be made more sophisticated while remaining transparent and easy to experiment with. The Tekkotsu software and curriculum materials are used by a number of universities, as well as in my own course here at CMU. Observing how students actually use the software allows us to refine it and explore new approaches to high level robot programming.

Teaching Children Programming Via Idioms

In computer programming, idioms are familiar bits of code that perform common or important functions. A programming language's level of abstraction determines the kinds of idioms it affords. "Incrementing a counter" is a low level idiom common to most languages. "Pursue and consume" is an important idiom in Kodu, a programming language designed for young children that provides high level primitives inspired by behavioral-based robotics. I'm working on developing a Kodu-based curriculum for K-6 students that explicitly teaches idioms and also places a heavy emphasis on finite state machines. Acquiring the concept of state machines at an early age is an important step toward computational thinking, and can equip a child to master a wide range of technologies.

Creating Affordable Mobile Manipulators

One of the obstacles to improving undergraduate robotics education has been the lack of affordable mobile manipulators. I have been developing mobile robots with serious vision and manipulation capabilities that can be used in a classroom. The first was the Chiara hexapod, which went on to play chess at AAAI and even play a bit of Bach on YouTube. More recently, the Calliope series of robots combine a wheeled base with a camera on a pan/tilt mount, an arm with a gripper, and a Linux netbook. The latest model, the Calliope3, has a target price below $1,000.