Robotic Dexirty Hand [C++, ROS2]

• Led software architecture on ARM Cortex-M7 within a cross-functional ME/EE/CS team, designing control systems for an 8-DOF robotic hand with wrist, power grasping finger, and dexterous finger for manipulation tasks
• Characterized serial elastic actuator (SEA) for force sensing and backdrivability across multiple loading conditions on all modules, achieving <0.05s force settling time and >10Hz bandwidth in position tracking
• Coordinating with eletrical team with motor driver board development integrating gate drivers, GaNFETs hall-effect sensors, and a 16-bit external ADC for current sense, achieving a more precise torque regulation
• Developed comprehensive brushless DC (BLDC) motion control libraries integrating forward/inverse kinematics with null space optimization algorithms, implementing inner loop with field-oriented control and PD torque control and outer loop PID position control and gravity compensation
• Architected inter-board (SPI & CAN) between 9 boards with sensors with Saleae logic analyzer for signal integrity verification, enabling 10kHz real-time feedback loops for encoders, current sensors, and motor drivers
• Created intuitive MediaPipe-based hand tracking graphical user interface with ROS2 integration, enabling real-time gesture recognition and visualization RViz with URDF-based digital twin implementation

Exploding Wild-Kitten Boardgame App [Java]

• Spearheaded implementation of test-driven development (TDD) for a Java-based multiplayer card game, creating comprehensive unit and integration tests that achieved 100% code coverage while applying boundary value analysis (BVA) techniques to identify and mitigate critical edge cases that could impact gameplay integrity
• Architected and deployed a robust CI/CD pipeline with strict branch protection rules that enforced automated quality checks, while establishing team-wide coding standards that improved maintainability and enabled successful internationalization (i18n) support across multiple languages

KUKA youBot Pick and Place [Python]

• Created a robotic manipulation simulation for KUKA youBot (4-wheeled omnidirectional mobile base + 5R arm), implementing trajectory generation using screw theory, SE(3) transformations, and Mecanum wheel kinematics
• Built PI feedback control system with pseudoinverse Jacobian for end-effector tracking, incorporating joint limit constraints and singularity handling for stable pick-and-place operations
• Designed modular codebase with configuration management and comprehensive error visualization, enabling autonomous manipulation tasks with error convergence across test scenarios
• View on GitHub

Vision Pro Simulator [Python, TensorFlow]

• Developed lightweight Vision Pro simulator with real-time hand gesture and motion recognition achieving >90% accuracy using custom MLP and LSTM neural networks, trained under 90 seconds on CPU
• Implemented interactive photo manipulation features including drag-and-drop, scaling, rotation, and drawing tools using computer vision techniques and bilinear interpolation
• Devised robust hand tracking system maintaining accuracy across varied lighting conditions and multi-user scenarios, with modular pipeline for detection, tracking, and gesture recognition
• View on GitHub
• See full video on YouTube

Invariant Unscented Kalman Filter for Distributed Robotic System [Matrix Lie Group, Multi-agent SLAM, Sensor Fusion]

• Proposed a target tracking algorithm for distributed time-variant system in 3D with invariant (using matrix Lie group and Lie algebra) UKF for distributed state estimation
• Applied Inverse Covariance Intersection sensor fusion algorithm as a consistent and less conservative method while handling unknown distributed covariance
• Performed 50 trials of Monte-Carlo simulation at different communication rate (including centralized system) and calculated position and orientation RMSE to show algorithm is robust to time-varying communication topology

Search Algorithm Visualization [C++]

• Created a visualization tool for search algorithm including A*, Dijkstra and BFS with different types of distance measurement (Euclidean, Manhattan, Chebyshev)
• View on GitHub

Planning and Control based on Vehicle Dynamics [C++]

• Implemented a Planning and Control algorithm based on vehicle dynamics to simulate and analyze vehicle behavior in complex driving scenarios such as overtaking, navigating crosswalks, and executing precise parking maneuvers
• View on GitHub

Lower Limb Exoskeleton Transparency Control [Python, ROS, Testing]

• Developed a test system for transparent haptics rendering on lower-limb exoskeleton using ROS Noetic
• Evaluated and validated parameters for PID and IIR filter by generating Bode plot and shift in RMSE in Python
• Achieved near zero output at human movement frequency with a less than 10ms delay in real-time

Line Following Robot [Embedded C++, PID]

• Devised a path following robot based on ATmega32U4 for at 4m/s in low lighting condition
• Tuned PID position controller based on IR sensors and PD speed controller based on encoder via trial-and-error, and attained an over 95% success rate
• Established UART communication with ESP32 as a Wi-Fi module to receive command from a higher-level controller, integrated a circuit consisting of a power source and LEDs for computer vision

2R Robot Impact Simulation [Lagragian Mechanics, Numerical Analysis, Python]

• Modeled a rigid body (with elastic impact) inside a box handled by a 2R robot as an end effector via Lagrangian mechanics and RK4, and implemented gravity compensation due to mass of the links and joints of the robot with optimal control
• View on GitHub

–>