Robotics

WATonomous - Robotic Arm

Developed an end-to-end Imitation Learning (IL) pipeline and camera-based teleoperation for a humanoid robot arm.

Imitation Learning: End-to-end IL pipeline on a simulated Franka Panda arm in IsaacLab; Differential IK teleoperation with keyboard control, 3×640×480 RGB capture, and GUI-triggered episode recording; exported robomimic-compatible HDF5 demos and trained policy using NVIDIA Isaac GR00T.
Camera Teleoperation: Real-time MediaPipe hand-tracking for a 21-DOF arm; DexRetargeting IK with spherical wrist decomposition; 7-frame median + EMA + jerk-limiter pipeline for stable low-latency 3D control from 2D coordinates; 3-camera sync recording saving HDF5+MP4 for GR00T fine-tuning.

PythonIsaacLabIsaac GR00TPyTorchDexRetargetingOpenCVHDF5ROS2Imitation Learning

Containerized autonomous navigation stack with global path planning and smooth trajectory following.

A* Search: Implemented A* search over grid-based costmaps for global path planning with obstacle avoidance.
Pure Pursuit Controller: Designed a Pure Pursuit controller for dynamic steering angle and velocity computation, enabling smooth and precise trajectory following.
Containerized Stack: Packaged the full autonomous navigation system (ROS2, C++, Docker/Gazebo) into a reproducible containerized environment.

C++ROS2DockerGazeboA*Pure Pursuit

Contributed to motion tracking improvements in humanoid robotics by exploring sensor technologies and Omnigraph integration.

Researched LiDAR sensors to enhance motion tracking precision and environmental awareness.
Explored Omnigraph modules to improve real-time responsiveness of humanoid motion systems.
Investigated integration strategies for combining sensor data with humanoid control systems.

IsaacLabLiDAROmnigraphMotion TrackingRobotics

Optimized and implemented advanced control strategies for autonomous vehicle systems.

Optimized PID control systems in Simulink, improving stability and precision in dynamic environments.
Implemented Gain Scheduling strategies to enhance accuracy across varying speeds and terrains.
Validated system performance using RoadRunner simulations to ensure real-world effectiveness.

PID ControlSimulinkRoadRunnerGain Scheduling

Designed and optimized autonomous systems and mechanical integrations for VEX competitions.

Developed bot algorithms and autonomous system functionality.
Designed and optimized the flywheel and sensor integration for improved performance.
Authored detailed engineering documentation to streamline collaboration and ensure high standards.
Contributed to the CAD model development of the VEX robot.

PythonSensorsMotorsPID ControlCADOnshape

Completed training in MATLAB's Simulink for control system design, modeling, and simulation.