My interests lie in the areas of Robotics, Computer Vision, Machine Learning, and AI. I want to make robots more general-purpose and robust; my research seeks to enable robots to move out of the lab and into real-world settings by bridging the gap between non-embodied AI and the noisy physical world.
My primary (thesis) work is in the area of robotic object perception. I developed an object representation, Bayesian Eigenobjects (BEOs), which enables robots to reason about novel objects in their environments in a rich and unified way. See more about BEOs here.
A colleague and I developed a system for combining pre-planned robot-motion trajectories with real-time human feedback to create a hybrid motion controller that allows a robot be responsive to human guidance. The project page has more information as well as a video demo of the system being used to facilitate a human-robot handshake.
I developed a method for learning from demonstration, DM-IRL, which allows a robot to learn from multiple demonstrations, with no requirements or assumptions on the quality of each demonstration, by leveraging an expert human judge. Please see the DM-IRL page for more details.
I was part of a group at UW-Madison focused using machine learning to combat online bullying. Our team created a system to detect and analyze bullying on twitter; among other thigs, our system attempts to determine if the bully later expressed regret over their actions. The data produced by our work is designed to improve the understanding, intervention, and policy-making on bullying. Please see the project page for more information.