Tabitha Oanda

Robotics Engineer · Fashion Designer · PhD Researcher at Brown University

Cloth is deformable, slippery, and hard to track — which means off-the-shelf robot setups don't cut it. I build the full stack: a bimanual hardware platform for manipulation, multi-camera perception using foundation vision models to detect and segment fabric across frames in 3D. I use teleoperation tools that make collecting training data practical. Following data collection, I train cloth dynamics models that can be used for model predictive control (MPC) and reinforcement learning policies for complex tasks.

My research is on getting robots to handle fabric reliably. That requires building the whole stack:

• Platform: custom aluminum frame, overhead-mounted bimanual arms, force-sensing silicone grippers
• Perception: multi-camera RGB-D, object detection and segmentation, 3D point tracking
• Data collection: leader-follower teleoperation for imitation learning, custom UMI-inspired data collection gripper
• Learning: cloth dynamics models trained and extended on data from my custom setup

A consistent theme in my research is taking methods developed in academic settings and adapting them to work on real cloth manipulation problems — the kind of contact-rich, deformable-object tasks that matter for industrial textile handling.