My Projects

Projects

Obstacle Avoidance using Control Barrier Functions

This project uses Control Barrier Functions (CBFs) to identify and avoid obstacles. The approach is implemented on a Go2 quadruped robot and AprilTags are used to detect obstacles.

Comparative Analysis of MPC, LQR, and RL-Based Footstep Planning Controllers in Uneven terrains

This project conducts a comparative analysis of Model Predictive Control (MPC), Linear Quadratic Regulator (LQR) and Reinforcement Learning footstep planners for bipedal robot locomotion in uneven terrains. The different control methodologies are evaluated with velocity tracking and success rates for different inital conditions and varying terrains inside a Drake simulation environment.

Optimization-based Estimation of Obstacles from Human Demonstration using Control Lyapunov Function and Control Barrier Functions

This project develops a obstacle identification method using Control Lyapunov Function (CLF) and Control Barrier Function (CBF). The approach probabilistically learns obstacle parameters from expert demonstrations, enabling robots to understand environmental obstacles without direct sensing.

Check out the poster and paper.

Pick-and-Stack Competition with vision-aided 7-DOF Manipulator Inverse-Kinematics Control

This project develops a vision-based pick-and-stack algorithm using the Franka Emika Panda 7-DOF manipulator. The block detection system uses real-time camera feedback from AprilTags and a gradient-based inverse Inverse-Kinematic optimization method was used for end-effector positioning. We designed waypoint trajectories to ensure collision-free manipulation.

Mobile Wheeled-Robot for Autonomous Navigation

The project designs a mobile robot platform capable of autonomous navigation through sensor integration and feedback control. The system includes a Time-of-Flight (ToF) sensor for distance sensing, a PID controller for stable motor control, an infrared (IF) sensor to detect beacons and a Vive sensor for localization. The robot can also be controlled with a webpage controller using ESP NOW.

Design of a 4-DOF Robotic Arm using Hydraulic control

This project involves the design of a 4-degree-of-freedom (DOF) manipulator using Fusion 360 and 3D printing and a controller driven by a modified unipolar step motor for higher torque. The system is fluid-actuated for smoother motion control and heat transfer fins are used to manage motor cooling and prevent overheating.

X-Corps Research Festival

The project creates a mobile application platform coupled with a solar-tracking controller to optimize panel efficiency. Using 11,278 Korean aerial images and semantic segmentation, we implemented an algorithm for rooftop solar panel placement planning. The projcet further incorporated predictive modeling using Korea Electric Power Corporation (KEPCO) metadata to forecast energy supply, demand, and pricing dynamics.

CDIC Competition

This project develops an AI-based surveillance application platform for child safety in daycare environments through real-time abuse detection. The project implements a multi-modal approach combining video detection that uses background suppression and the frame differences to isolate motion patterns from real CCTV footage, and acoustic event classification using Mel-Frequency Cepstral Coefficient (MFCC) features extracted from audio.