Mechatronic For Our Future Life

CoroBots are broadly capable and expandible four-wheeled mobile robot platforms designed to minimize the cost and complexity of robotics research and development. When Robot learned that CoroBots and Microsoft Robotics Studio were in active use at Vassar College, we made a beeline for Ken Livingston, Professor of Psychology and Cognitive Sciences, to give us an insight into what he and his students are up to. We then interviewed CoroWare to learn more about their history and mission in educational and mobile robotics.
—Tom Atwood, Editor-in-Chief, Robot Magazine

The Interdisciplinary Robotics Research Laboratory was funded by a grant from the National Science Foundation, with matching funds from Vassar College, to provide a space in which faculty and students from many different departments might pursue research in robotics. Lead faculty on the grant were from departments of Biology (Jon Long), Computer Science (Tom Ellman, Luke Hunsberger, Brad Richards), and Psychology (Ken Livingston), and all had worked together previously in Vassar's Cognitive Science Program. There are more than a dozen different robots in the lab, but CoroWare has provided our core suite of machines. The platforms are very flexible, with easy access to components and the ability to add large numbers of sensors in many different configurations. We've used everything from simple tactile sensors to fairly high resolution cameras. This allows us to use the robots in many different research projects.

In addition, CoroWare has designed a software interface that makes communication with the robots very easy to manage, whether they are being teleoperated or are running autonomously. We became very early adopters of Microsoft Robotics Studio as our primary software platform because it allows us to integrate code written for specific purposes in different languages. For example, our vision code is primarily written in C, but other functions are handled in C++ or C#. MSRS does the work of making these different chunks of code talk to each other seamlessly. CoroWare has integrated a very power efficient electronics package (built around Via boards and chips) with plenty of computing power, so we can easily take advantage of MSRS even when running computationally complex algorithms.
The fact that the same robots can be used in teleoperated or in autonomous mode with no major hardware changes has been a real boon for us. We can run our experiments on how variation in sensor input affects controllability in a teleoperation task one day and the next day use the same robot in our experiments that explore autonomous category learning. The fact that we can compare data from human teleoperators with data from autonomous robots turns out to be a real plus. For example, our experiments are already yielding some interesting and counter-intuitive results, including the finding that sometimes perceiving the world with less detail and precision makes effective learning much easier, not harder.