Portfolio Presentation

Education

Seoul National University of Science and Technology (SEOULTECH)
M.S., Electrical & Information Engineering — GPA 4.19 / 4.50
Seoul, South Korea · Feb 2024 – Aug 2026 (Expected) · Research Assistant
Universitas Indonesia
B.S., Electrical Engineering — GPA 3.61 / 4.00
Depok, Indonesia · Aug 2019 – Aug 2023

Technical Skills

Focus: Robotics (mobile robots) · Control & Estimation · Electrical Engineering

Control & Estimation: MPC / MPPI / MPCC, PID, LQR, Kalman Filter (KF/EKF), 2D SLAM, Motion Planning & Trajectory Optimization
Machine Learning: Gaussian Process (GP/SGP), Neural Networks, Imitation Learning, PyTorch & TensorFlow (basic)
Programming & Frameworks: ROS1 / ROS2, C/C++, Python, MATLAB/Simulink, CasADi, Flutter, HTML/CSS/JS, OpenCV (basic)
Tools & Platforms: Gazebo, move_base, Linux, Git, Docker
Hardware & IoT: NVIDIA Jetson, Raspberry Pi, Arduino, ESP32

Performance-Enhanced Risk-Aware MPPI using Gaussian Process

Objective

Drive a robot safely despite the model–reality gap, without hand-tuning the controller's safety level.

Outcome / Contribution

A Gaussian Process learns the model gap offline and sets the safety level automatically (risk-aware MPPI) — no manual tuning.
On F1TENTH (sim + real): finished every obstacle lap where baselines crashed, at the same speed.

Technical skills

Gaussian Process MPPI / sampling MPC Localization & mapping (SLAM) ROS1 move_base (Nav2 equivalent) Gazebo PyTorch PyCUDA Python C/C++

Hardware: Jetson Orin · LiDAR · IMU · wheel odometry · VESC

Real-world run — proposed method

Method framework: offline data collection, Gaussian Process, and online risk-aware MPPI control — Method framework

The F1TENTH autonomous car platform with Hokuyo LiDAR, Jetson Orin NX, VESC, and IMU — F1TENTH platform

Simulation trajectory with obstacles — Simulation — with obstacles

Simulation trajectory without obstacles — Simulation — without obstacles

Images, video & explanation → Paper (IEEE Access, 1st author) →

Robust RA-MPPI using Online Learning

Objective

Adapt to uncertainty in real time — learn the model gap while driving, with no data collected beforehand.

Outcome / Contribution

A Sparse Gaussian Process learns the gap online in the background, never slowing the controller.
On F1TENTH (sim, with obstacles): 22 laps vs. 16 for the previous method, still real time.

Technical skills

Sparse Gaussian Process (online learning) MPPI / sampling MPC Localization & mapping (SLAM) ROS1 move_base (Nav2 equivalent) Gazebo PyTorch PyCUDA Python C/C++

Hardware: Jetson Orin · LiDAR · IMU · wheel odometry · VESC

Simulation (Gazebo)

Real-world run — proposed method

Control scheme: a Sparse Gaussian Process feeding a risk-aware MPPI controller in the loop — Control scheme

Online trajectory sampling around the vehicle — Online sampling

Images, video & explanation →

1st author · Submitted to IJCAS (Springer)

ESP32 IoT Medical Gateway

Objective

Connect medical instruments to the cloud for posyandu (community health post) measurements.

Outcome / Contribution

Built an ESP32 gateway linking thermometers, oximeters, and biometric sensors.
Connected over six categories of medical devices.

Technical skills

Embedded / IoT RTOS (FreeRTOS) BLE connectivity Reverse engineering C/C++

Hardware: ESP32

Internship at DoctorTool · Team of 2 · 2022

IoT medical gateway with connected thermometers, oximeter, scale, and blood-pressure monitor

6-DOF Robot Manipulator Simulation (OpenGL)

Objective

Visualize and control a 6-DOF robot arm in real time.

Outcome / Contribution

Real-time 3D arm in C/OpenGL (GLUT) with forward and Jacobian-pseudoinverse inverse kinematics.
Point-to-point trajectories, joint commands streamed over serial.

Technical skills

C OpenGL / GLUT Robot kinematics (FK/IK) MATLAB Robotics Toolbox (Peter Corke, MATLAB)

Course project · Universitas Indonesia · May 2023

Source →

Slides (PDF) →