Felix Duvallet

PhD, Robotics

I am now a postdoctoral fellow at the Ecole Polytechnique Federale de Lausanne (EPFL). More info coming soon.

I completed my PhD in Robotics at Carnegie Mellon University. My research focuses on applying imitation learning techniques to difficult decision-making problems in robotics. I'm primarily interested in problems of human-robot interaction, my recent work is on enabling robots to understand natural language in unknown environments.

For example, the wheelchair robot in the video below is instructed to "go to the cone that is behind the trashcan." Our approach to solving this problem includes hypothesizing a distribution of maps, and learning a policy that reasons about uncertainty.

My advisor was Tony Stentz. During my PhD, I also worked closely with Tom Kollar, Sachi Hemachandra, Matthew Walter, and Tom Howard.

I successfully defended my thesis on natural language direction following for robots in unstructured unknown environments in January of 2015. My thesis committee consists of Tony Stentz, Drew Bagnell, Manuela Veloso, and Nick Roy (MIT). In 2013, I was selected as an HRI Pioneer.

Previously during my PhD, I've also studied multi-robot task allocation for complex tasks, specifically using imitation learning to learn the utility function used during task allocation. Before graduate school, I spent a year in Australia at the CSIRO Autonomous Systems Lab, working on Wifi localization for industrial environments.

In my spare time, I climb rocks of all sizes (from boulders to mountains), ride motorcycles and bicycles, and drink coffee.

Below is a cloud of words from my dissertation: