Skip to content
View mincasurong's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report mincasurong

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
mincasurong/README.md

Hi, We're M9G 👋

From Machine to General Intelligence by Agent-enabled System Engineering

Minsu (Max) Chang, Ph.D. • Robotics Research Engineer at Samsung Electronics (SAIT)
Spatial Intelligence, Physical AI, Multi-Robot Task & Motion Planning (TAMP), and Rehabilitation Robotics


Main Portal CV / vCard Research Blog Research Archive

Google Scholar LinkedIn GitHub


🏆 Recent News • 📝 Technical Research Blog • 🤖 Gemini × Robotic Arms • 📚 Publications (21 Papers & 14+ Patents) • 📱 Free Live Apps • 🎓 Tutorials ↗


🏆 Recent News & Highlights

🥇 1st-Place Winner at Google Gemini 3 Seoul Hackathon 2026

Claimed 1st Place at the prestigious Google Gemini 3 Seoul Hackathon 2026, standing out among 111 finalist projects and over 1,500 highly competitive applicants.

Google Gemini 3 Seoul Hackathon 1st Place - GeminiSpace Demo Video

▶️ Click to watch: GeminiSpace Demo Video (Google Gemini 3 Seoul Hackathon 1st Place)

About GeminiSpace:
We developed GeminiSpace, a Vision-Language-Action (VLA) navigation system that turns standard 360° indoor room photos into an interactive indoor semantic map. Simply capture your space and query it naturally, acting as a smart GPS for any indoor environment.

Live Web App Google Blog Winners Gallery GitHub Repo


📝 Featured Technical Research Blogs (with 40s Video & TL;DR)

Explore all visual TL;DRs and 40-second gemini-omni-1.1-flash video syntheses at mincasurong.github.io/blog:

M9G-TR-2026-05: UIUC Lazy-DaSH (IEEE T-RO) x IsaacSim_Gemini

M9G-TR-2026-05 • IEEE T-RO (UIUC) • 16 Arms / 60+ Obj • 100× Faster

Summarizing Seongwon "Scott" Lee et al.'s Lazy-DaSH (UIUC Parasol Lab, arXiv:2504.05552) and mapping its 4-layer lazy hypergraph (H_T → H_M) & branch-scoped constraint feedback onto Gemini 4 Argon, Gemini Robotics-ER 2, and Gemini 3.5 Flash-Lite with 7 continuous KPIs.

Read Lazy-DaSH Blog & Watch 40s Video ↗

M9G-TR-2026-04: SOTA Survey on Multi-Robot Assembly, Deformables & SDLs

M9G-TR-2026-04 • 14 SOTA Systems • Assembly, Deformables & SDLs

4-tier neuro-symbolic + GPU physics synthesis across MATTERIX, APEX-MR, RoboFactory, DexGarmentLab, ORGANA, and Project STAMP (v1.0.0-stamp).

Read SDL Survey & Watch 40s Video ↗

M9G-TR-2026-03: Project STAMP Deterministic Safety Gate

M9G-TR-2026-03 • 100% Fault Shield (610/610) • 1,000 Episodes

4-stage closed-loop Spatial, Task & Motion Planning (Plan → Gate → Execute → Write-Back) intercepting 610/610 kinematic, AABB collision, torque, and zone-lock faults in Isaac Sim 6.0 & ROS 2 Jazzy.

Read STAMP Blog & Watch 40s Video ↗

M9G-TR-2026-02: m9g-agent-corps to m9g-mcp

M9G-TR-2026-02 • 28 Modules / 169 Tests • Pivot to m9g-mcp

Source-level post-mortem of engineering a 28-module coding harness, its 6 standalone failure modes, and extracting its 7 deterministic tools into the m9g-mcp server.

Read Agent Blog & Watch 40s Video ↗

M9G-TR-2026-01: Symbolic 3D /tf Grounding & Pipelined VLA

M9G-TR-2026-01 • 3-Arm FR3 Tower • -76.6% API Cost ($0.01972)

3-Stage hardware-in-the-loop benchmark in Isaac Sim 6.0 & ROS 2 Jazzy achieving a 10-turn pipelined schedule with 0 RGB tokens and -76.6% API cost ($0.01972).

Read VLA Blog & Watch Side-by-Side Video ↗


🤖 Gemini × Robotic Arm Series (Isaac Sim 6.0 & ROS 2 Jazzy)

Multi-Robot VLA Manipulation, React Flow Digital Twins & Hardware-in-the-Loop Benchmarks (mincasurong/IsaacSim_GeminiRobotics).

Gemini Robotics ER-2 vs. Gemini 4 Argon Benchmark

Side-by-side 3-arm 9-block tower benchmark in Isaac Sim 6.0: software-pipelined 10-turn scheduling and hybrid Gemini 4 Argon + Flash-Lite cutting API cost by 76.6% ($0.01972) with 0 RGB tokens.

EP 12. Multi Robot Collaboration

Three FR3 robotic arms cooperatively constructing geometric formations in Isaac Sim with dual zero-token React Flow DAG and Gemini Robotics-ER 2 VLA execution.

EP 11. FR3 Robots Cooperating with Gemini Robotics-ER2

Cooperative FR3 robotic arms planning and assembling in NVIDIA Isaac Sim via Gemini multi-agent reasoning and React Flow digital twin.

EP 10. Natural Language Input - Isaac Sim, ROS 2, Gemini

End-to-end VLA multi-robot control framework integrating Gemini Robotics-ER 2, Isaac Sim, ROS 2, and real-time Web monitoring GUI.


📚 Publications & Research Archive

Full list of 21 Peer-Reviewed Publications and 14+ US/KR Patents is available at mincasurong.github.io/archive and mincasurong.github.io/vcard.

Year Venue Paper Title Link
2026 IEEE/RSJ IROS SPLiT: Spatially Generalizable Phase-Partitioned Motion Generation for Adaptive Liquid Transfer (H. Kim, S. Jeon, M. Chang, H. Lee, J. Park) Google Scholar ↗
2025 IEEE ICRA OPPA: Online Planner's Parameter Adaptation for Enhanced Mobile Robot Navigation IEEE Xplore ↗
2020 IEEE TNSRE AI Therapist realizing Expert Verbal Cues for Effective Robot-Assisted Gait Training IEEE Xplore ↗
2019 IEEE TNSRE Research on Terrain Identification of the Smart Prosthetic Ankle by Fuzzy Logic IEEE Xplore ↗
2019 Applied Sciences Precise flowrate control of fluid gear pump in automated painting system using a repetitive controller MDPI ↗
2017 ICORR / Rehabweek A research on the postural stability of a person wearing the lower limb exoskeletal robot by the HAT model IEEE Xplore ↗

🦿 Virtual Reality (VR) + Wearable Robots for Rehabilitation (Click to expand/collapse)
VR Lower Limb Exoskeleton
VR-Integrated Lower Limb Exoskeleton
Interactive VR environment with synchronized biofeedback for gait rehabilitation.
Adaptive Impedance Control
Adaptive Impedance Control Exoskeleton
Gamified assist-as-needed gait training using adaptive impedance control.
Virtual Spring Damper
Virtual Spring-Damper Gait Guidance
Trajectory constraint control along predefined paths for robotic gait assistance.
🚙 Mobile Robots, Navigation & Manipulators (Click to expand/collapse)
BARN Benchmark
BARN Benchmark Path Planning
Autonomous mobile navigation and dynamic obstacle avoidance benchmarking.
ROS 2 & MoveIt 2
ROS 2 & MoveIt 2 Motion Planning
Collision-free robotic arm trajectory planning running natively on Windows 10.
Jetson Nano ROS Autopilot
Jetson Nano ROS Autopilot
Edge AI autopilot and sensor-driven motion control on Jetbot platform.

📱 Live Deployed Apps & Tools (100% Free Access)

UrKuma: Korean-Japanese Quiz & learning

Master Chinese characters and Japanese vocabulary in 3 minutes a day on Google Play.

OPIc AI Master Coach

Real-time conversational AI voice sparring trainer for OPIc Advanced Low (AL).

JLPT N4 Mock Exam & Drill Master

Full-section Japanese proficiency exam simulator with 1:1 Gemini AI weak-spot tutoring.


🛠️ Core Tech Stack & Frameworks

Domain Technologies
Robotics & Simulation NVIDIA Isaac Sim 6.0 ROS 2 Jazzy MoveIt 2 MuJoCo 3.x Franka Emika FR3 Jetson Nano
AI & Spatial Intelligence Gemini 4 Argon Gemini Robotics-ER 2 Gemini 3.8 / 3.5 Flash-Lite Visual SLAM VLA & TAMP
Full-Stack & Digital Twins React 19 TypeScript 5.8 React Flow v12 Vite 6 Tailwind CSS v4 FastDDS Unicast
Cloud & Agentic DevOps Google Cloud Run Model Context Protocol (MCP) Docker GitHub Actions PyTorch

The M9G Formula: m incasuron g (Human Engineering) × m insu chan g (Agent-enabled Intelligence)

Portal • CV • Blog • Archive

© 2026 M9G (Minsu Chang, Ph.D.) • Built in Public • mincasurong.github.io

Pinned Loading

  1. GeminiSeoulHackathon2026 GeminiSeoulHackathon2026 Public

    Python 31 7

  2. IsaacSim_GeminiRobotics IsaacSim_GeminiRobotics Public

    Python 47 3

  3. Admittance-Control-Simulink Admittance-Control-Simulink Public

    This storage includes simulink programs of the admittance control and impedance control

    11

  4. Online-LQR Online-LQR Public

    This storage is related to the Online LQR which is combination between LQR and reinforcement learning

    MATLAB 6 2