Robotics Engineer — LiDAR SLAM · Autonomous Navigation · Multi-Agent Fleet Systems

I build autonomous mobile robot systems end to end — from LiDAR-inertial SLAM and localization to fleet-level coordination. I implement SLAM front-ends and back-ends from the equations up — ESIKF, on-manifold optimization, analytically derived ICP Jacobians — and I'm currently working on degeneracy-aware LiDAR-inertial odometry for solid-state sensors.

In industry since 2022: 30+ AMRs deployed across a 23,000 m² logistics center, precision drilling automation for aerospace manufacturing, and a handheld 3D mapping device built from the ground up.

Based in Daejeon, South Korea.

Handheld 3D mapping device with real-time LiDAR SLAM for large-scale environments — custom hardware and the full software pipeline. The localization core is open-sourced as esikf-lio-core below.

▶ Watch it in action — real-time mapping demo

Web-based 3D point cloud viewer and analysis tool. No install, runs in the browser — LAS, LAZ, PLY, PCD, XYZ, PTS.

The localization core of HandheldSLAM, open-sourced — an iterated error-state Kalman filter (18-dim state, on-manifold update, motion deskewing) written from the equations up, with the full derivation in the README. Ships a real cross-session test: a Livox Mid-360 walk localized against a map built from a different session — 3.4 cm ATE RMSE out of the box, reproducible in five commands.

git clone https://github.com/warhammer50K/esikf-lio-core &&cd esikf-lio-core
cmake -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build -j
tools/get_example_data.sh # real Mid-360 recording + map (~135 MB)
./build/esikf_lio_core example_data/params.json # → trajectory.txt
python3 tools/plot_trajectory.py trajectory.txt \
--map example_data/esikf-lio-example/map.ply \
--gt example_data/esikf-lio-example/ref_keyframes.txt # → trajectory.png + ATE

IMU noise characterization via Allan variance — extracts noise density and bias instability parameters for SLAM frameworks (GTSAM, LIO-SAM, FAST-LIO).

Real-time object segmentation with Meta SAM 2.1 on an OAK-D Lite, served via FastAPI.

FLIR Blackfly GigE camera viewer built on the Spinnaker SDK, C++17.

Single-header, dependency-free C++11 library for exchanging plain structs over UDP — typed send/receive, framing, background receive thread. Born from joystick-to-robot teleoperation.

• Warehouse AMR fleet — 30+ robots in a 23,000 m² logistics center: SLAM-based localization, autonomous navigation, multi-agent path finding, and the fleet management system coordinating them.
• Aerospace manufacturing — AMR control for precision drilling automation.
• 3D vision — structured light, depth camera calibration, real-time segmentation, robotic shoe bonding.
• Platforms — serving robots, AMR, mecanum-wheel platforms, AMMR (mobile manipulation).

pixvnt151@gmail.com · YouTube · Daejeon, South Korea