I build robotic systems that stay operational in the real world and I research architectures where robots become legible teammates (not opaque tools).

Links

• Blog: https://jack0682.github.io/
• GitHub: https://github.com/jack0682
• Email: jaehongoh1554@gmail.com

Human-centered, safety-critical robotics under partial observability

• Semantic perception → mapping → decision → explainable action
• Topology-conditioned reasoning (ONN / ORTSF)
• Constraint-aware RL for manipulation (Doosan M0609)

Core thesis: Robust autonomy = semantic continuity + safe decision-making + operational loops.
Last updated: 2026-02-12 (KST)

• Nav2 tuning for real environments (local/global planners, costmaps, recovery behaviors)
• Reproducible failure triage with rosbag + logs (issue isolation → fix → regression test)
• Real-time constraints mindset: latency budgets, QoS, callback scheduling

• Detector/tracker pipeline for safety events (e.g., PPE / hazards)
• Event triggers → robot behavior (stop/alert/avoid) + logging
• Operator-facing monitoring via MQTT → WebSocket → Web UI

• Scene-graph / ontology-ready state representations
• Topology/constraint framing for stability and reasoning (ONN / ORTSF)
• “Why did the robot do that?” → human-readable rationales

• Ontology Neural Networks for Topologically Conditioned Constraint Satisfaction — arXiv:2601.05304 (2026)
  https://arxiv.org/abs/2601.05304
• Constructive Lyapunov Functions via Topology-Preserving Neural Networks — arXiv:2510.24730 (2025)
  https://arxiv.org/abs/2510.24730
• Ontology Neural Network and ORTSF: A Framework for Topological Reasoning and Delay-Robust Control — arXiv:2506.19277 (2025)
  https://arxiv.org/abs/2506.19277

• Cognitive Synergy Architecture: SEGO for Human-Centric Collaborative Robots — arXiv:2506.13149 (2025)
  https://arxiv.org/abs/2506.13149
• Towards Cognitive Collaborative Robots: Semantic-Level Integration and Explainable Control for Human-Centric Cooperation — arXiv:2505.03815 (2025)
  https://arxiv.org/abs/2505.03815

• Learning to Assemble the Soma Cube with Legal-Action Masked DQN and Safe ZYZ Regrasp on a Doosan M0609 — arXiv:2508.21272 (2025)
  https://arxiv.org/abs/2508.21272

• Repo: CSAv1 — https://github.com/jack0682/CSAv1
• Keywords: semantic mapping, scene graph, explainable control, HRI-ready interfaces

• Repo: ONN — https://github.com/jack0682/ONN
• Anchored by: ONN/ORTSF papers above

• Repo: web_robot_interface — https://github.com/jack0682/web_robot_interface
• Purpose: command-center style monitoring (ROS2 Robot → MQTT Broker → WebSocket Bridge → Web Dashboard)

• Repo: Projection_plane — https://github.com/jack0682/Projection_plane

• Repo: linetracking — https://github.com/jack0682/linetracking

• ROS2 (Humble), Nav2, TF2, URDF, RViz
• rosbag-based debugging, QoS tuning, callback groups / executors
• Multi-robot topic/namespace design, message/service/action contract design

• RGB-D pipelines (depth → 3D points → map-frame transform)
• Detection / tracking integration (YOLO + tracking stack), event extraction
• SLAM / mapping: ORB-SLAM2, RTAB-Map
• Point cloud ops: PCL, registration/filters (typical: voxel, crop, plane/cluster)

• Classical control: PID, feedforward+feedback, state-space intuition
• RL: DQN variants / PPO usage patterns, action masking, safety guards
• Manipulation safety: constraint design, singularity-aware motion constraints (Doosan M0609 context)

• MQTT (robot telemetry + events), WebSocket streaming
• FastAPI-style backend patterns, structured logging, config-driven systems (YAML/JSON)
• Telemetry storage patterns (time-series mindset; reproducible logs)

• Python (primary), C/C++, MATLAB
• Git, Linux (Ubuntu), Docker/venv, CI-friendly repo structuring
• Data formats: JSON/CSV/JSONL; scene-graph export (Neo4j-ready structuring)

I’m interested in:

• human–robot teaming in safety-critical environments
• occlusion / partial observability + risk-aware planning
• explainable policies that can be audited in the field

Best contact: jaehongoh1554@gmail.com