Drone racing
For this project, I developed an RL-based autonomous drone racing policy with successful sim-to-real transfer, achieving 1st place out of 17 teams in a competitive class project. Trained policies in simulation using domain randomization to improve robustness to real-world dynamics and sensor noise.

Trajectory optimization with complementarity constraint
This project tackles the trajectory optimization problem involving a 2-link robotic gripper. The goal was to plan a feasible motion for the gripper to pick up an object and lift it to a specified height. The problem incorporates complementarity constraints to accurately model the contact interactions between the gripper and the object.

Hand gesture recognition
An acoustic device is attached to the wrist, emitting a fixed sound wave. As the hand moves, the received sound wave changes depending on the gesture. We use real-time machine learning to distinguish hand gesture based on the received sound wave.

Lane alignment autonomous driving
This is a robot car with basic self-driving ability. The car uses a RGB camera to detect the lane and obstacles. It follows the lane even the lane is partially obscured. When an obstacle is present, an animal in our case, the car overrides the autonomous driving system, crosses the lane with minimum deviation, and returns to the lane when it's safe.

3 Link biped
This video of a walking 3 link biped is simulated with MatLab