Robotics Projects

The Large Behavior Models project at TRI aims to build foundation models for dextrous robotic manipulation. I'm a research co-lead for the 3D Generalizable Representations (3DGR) subgroup, and have also co-led real-world large-scale hardware evaluations of multi-task generative manipulation models across the Los Altos and Cambridge offices.

The DCIST Collaborative Research Alliance (CRA) is a multi-university project. I worked on integrating our planner into the photo-realistic simulation stack, and explored new applications of our learned planning algorithm.

Under the SaRUC project, the MIT-NASA team investigated how a multi-agent team of autonomous quadrotors could be deployed to search for a lost hiker in a wooded area. I worked on the integration of the various substacks and sensors both in simulation and on hardware, and directed flight testing as the ground station manager during our experiments at the NASA Langley test site. We published the details of the system and experiments, and the project was one of the recipients of the 2019 AUVSI Humanitarian XCELLENCE Award.

The goal of the DARPA FLA project was to enable autonomous quad rotor flight up to speeds of 45 MPH in GPS-denied environments without prior maps. In my first year of grad school, I prototyped and tested the initial global planning algorithms, and supported flight testing at MIT sites and Joint Base Cape Cod with the MIT-Draper team.

For a few years in undergrad, I participated in the AUVSI sUAS competition project as a member of the UCSD airframe team. I got to do lots of neat things such as build a wet layup composite for a fixed-wing aircraft, write some simulation code for spar bending analysis, and go to Pax River as part of the competition team.