Robot Perception · 2026
SAGE: Semantic And Geometric Estimation of 6D Object Pose from Multi-View Observations
A multi-view transformer combining geometric consistency and open-vocabulary semantic grounding for real-time 6D object pose estimation.
Ph.D. Candidate in Robotics
Humanoid and legged-robot learning, real-time object perception, and vision-based control.
Robot Perception · 2026
A multi-view transformer combining geometric consistency and open-vocabulary semantic grounding for real-time 6D object pose estimation.
Robot Perception · 2026
Category-level pose, scale, and shape estimation formulated on SIM(3) and solved with a high-throughput matrix ADMM method.
Humanoid Robotics · 2026
Long-horizon humanoid box rearrangement through reusable skills and a shared, task-agnostic whole-body controller, validated in simulation and on Digit V3.
Robot Perception · 2025
Active shape models, ADMM registration, and multi-hypothesis tracking for robust object-centric perception in dynamic environments.
Legged Locomotion · 2025
A vision-based reinforcement-learning footstep planner paired with operational space control for Cassie locomotion over unstructured terrain.
Computer Vision
Stereo-camera terrain recognition for adaptive wearable assistance in dense forest environments.
Machine Learning
A real-time learning-based system for diagnosing gear damage in electric-vehicle reducers.
Computer Vision and Control
Real-time UAV detection and tracking that combines a one-stage detector with a Siamese network.