Brown University undergraduate Nishanth Kumar, a computer engineering concentrator and computer science researcher, has recently received the Barry Goldwater Scholarship for his research into Learning from Demonstration (LfD).
The scholarship was established by Congress in 1986 to identify and support the next generation of research leaders, and is widely regarded as one of the most prestigious undergraduate scholarships in the natural sciences, mathematics, and engineering in America. Nishanth joins fellow Brown students Adam Tropper (Physics and Astronomy concentrator) and Lucas Sanchez (Chemistry concentrator) this year as a scholarship recipient.
Nishanth’s research focuses on teaching robots to learn real-world skills directly from observing demonstrations. “LfD allows non-expert operators to program skills simply by demonstrating them many times,” he explains, “and these learned skills are more general: they are able to handle slight variations of a task, such as if an object to be placed is slightly misplaced.” The issue with current LfD techniques, however, is that they train skills that are unable to target specific goals from many possible choices (i.e. targeting a specific button within a grid) without copious amounts of training data.
“To combat this issue, I helped propose a method that learns skills that are parameterized by a goal parameter,” Nishanth says, “such that altering this parameter correctly alters the skill. In the button pressing scenario, instead of training a new skill for each button, we train one general skill that adapts itself depending on where the button is.”
Looking forward, Nishanth is ready to continue solving some of the most practical problems in Artificial Intelligence: “After winning the scholarship, I’ve felt a deep responsibility and motivation to continue my research into AI and robotics. I believe the advent of intelligent, collaborative robots can massively change the world for the better and I hope to play some part in making this dream a reality.”
- Brown CS