Improbable AI Lab
During my sophomore year, I began to research at the Improbable AI Lab at MIT under the supervision of Haoshu Fang. One of the largest bottlenecks in the robotics revolution is access to a dataset that can be used to train manipulators. The project aimed to address this issue by building an exoskeleton arm that could be used to gather manipulation data to train actuators through transfer learning.
Design
I modified the design of the fingertips as we sought to include a tiny endoscopic camera that would be able to gather tactile feedback alongside Gelsight technology. I also modified the textures of the fingertips to allow for better force feedback and adjusted the linkages to avoid collisions.
Build
I assisted in the assembly of the exoskeleton, 3D printed parts, and learned to use a metal laser cutter to cut the linkages. I also used a vertical mill, band saw, and belt sander to prepare the parts.
Results
This research opportunity reinforced my CAD skills in OnShape. It also furthered my machining skills as I gained more experience using a metal laser cutter and a vertical mill.