Stefanie Tellex, an assistant professor in the computer science department at Brown University, talked with Providence Business News about the work she does in the Humans to Robots lab, its recent Rethink Robotics Video Challenge win, which gave it its third Baxter robot, and the lab’s “million object challenge” involving Baxter robots.
PBN: What interested you in robotics as a career?
TELLEX: From when I first learned to program as a young child, I was interested in making computers that could talk. I moved to robotics because I realized to really be able to talk to people, a computer needed to know something about the physical world. Robots can change the physical world, so they can help people in all areas of life. The potential impact is staggering.
PBN: Can you explain what kind of work is done in the Humans to Robots lab?
TELLEX: We are at the cusp of a revolution in robotic technology that will fundamentally change how we live and work. Hospital robots will check on patients and report their status to nurses, saving time and improving patient outcomes. Childcare robots will help parents with chores such as assisting at diaper changing or feeding, so that families can spend high-quality time together. Manufacturing robots will collaborate with people to assemble complex objects on reconfigurable assembly lines, increasing the efficiency and flexibility of factory floors. The robotic revolution will approach and surpass the ubiquity and usefulness of the computer revolution, because robots can change not only the virtual world, but the physical world as well. The aim of our research program is to create robots that collaborate with people to meet their needs, so that human-robot collaboration approaches the ease of human-human collaboration. To create collaborative robots, we focus on three key challenges: 1) perceiving the world using the robot’s sensors; 2) communicating with people to understand their needs and how to meet them; and 3) acting to change the world in ways that meet people’s needs.
PBN: Congratulations on winning the Rethink Robotics Video Challenge. Was it difficult to program the robot so that it learned to pick up objects from experience?
TELLEX: Yes. We are working on new tools for interactively programming the robot that are quite powerful. They let you write “robot one-liners” to quickly create behaviors interactively with the command line. We use that framework for most of our research.
PBN: What will the lab do with the third Baxter robot?
TELLEX: We're not sure yet. We've talked about robot lightsaber battles. For sure we will use it for data collection – when a robot is running object-scanning jobs that take six to eight hours apiece, it’s really easy to use up robots. We’ve also been considering putting one on a mobile base.
PBN: How is the Million Object Challenge going? How did you get the idea for that?
TELLEX: At the beginning, I wanted a robot that could pick up about 10 different objects. Most labs achieve this by picking out a few specific objects and then manually adapting a system to those specific objects.
We realized though, that we could enable the robot to automatically adapt itself to the object. That way it wouldn’t require an expert to add new objects, but anyone could just hand an object to the robot and have it learn about that object. From there, we realized we could scale the whole process up, and do it automatically. And then we noticed that there were a lot of these Baxters around, and most of them weren’t being used a lot of the time.