R
AI-Readable
Manufacturing knowledge, control logic, process flows, and equipment states are represented in structured forms that AI systems can interpret.
POSTECH · Factory Intelligence Laboratory
AI-Native Manufacturing Systems
We develop manufacturing systems that are readable, reasonable, actionable, and learnable for AI.
Our research integrates control semantics, autonomous agents, digital twins, physics-aware simulation, domain-robust fault diagnosis, and causal quality analytics to build the next generation of intelligent and adaptive factories.
ReadAI-readable semantics
ReasonAI-reasonable models
ActAI-actionable control
LearnAI-learnable systems
Research ThemeAI-ready factories of the future
Vision
Toward manufacturing systems that AI can understand, reason about, act upon, and learn from.
Conventional manufacturing systems are primarily designed for human engineers: drawings, control logic, process rules, and operating procedures are often fragmented, implicit, and difficult for AI to interpret.
FILab transforms factories into AI-native environments by developing semantic representations, simulation models, optimization methods, digital twins, and learning algorithms that make manufacturing knowledge executable for AI agents.
AI-Native Properties
What makes manufacturing AI-native?
We make production systems structured enough for AI to read, model, execute, and improve them across machines, workcells, and factories.
R
AI-Reasonable
AI can infer operational constraints, causal relationships, control semantics, and process dependencies from manufacturing representations.
A
AI-Actionable
AI decisions can be translated into executable actions through digital twins, control programs, robotic systems, and automation interfaces.
L
AI-Learnable
Manufacturing systems continuously improve by learning from production data, fault patterns, quality outcomes, and human tacit knowledge.
Research Pillars
Two pillars for AI-native manufacturing
AI-Native Machine Control and Operations Autonomy
Control semantics architecture, autonomous workcell agents, digital twins as AI decision spaces, physics-aware virtual commissioning, and human tacit knowledge for robotic learning.
Factory ControlNext-Generation Fault Detection, Prognostics, and Causal Quality Analytics
Continual and source-free domain adaptation, open-set and target-free domain generalization, diffusion-assisted cross-domain fault diagnosis, and process-genealogy mining.
Factory AnalyticsSoftware in Action
AI-native tools for virtual programming and autonomous workcell operation.
Our software platforms demonstrate how manufacturing semantics, digital twins, and agent-based interfaces can turn workcell knowledge into executable control and operation workflows.
Virtual Programming Platform
MEL Studio for Virtual Programming
MEL Studio illustrates how manufacturing tasks, device states, and control semantics can be organized into a virtual programming environment for AI-native machine control.
Autonomous Workcell
Agent Builder Overview
Agent Builder supports the configuration of AI agents that connect devices, interpret workcell states, and prepare autonomous operation logic.
Agent Execution
Agent Builder: Configuration to Execution
This demo shows how agent configuration can be connected to execution workflows for AI-assisted workcell operation and decision-making.
Featured Topics
Control, diagnosis, quality, and design connected by AI.
Contact
Factory Intelligence Laboratory
Department of Industrial & Management Engineering, POSTECH
Tel: +82 54 279 8261 · Fax: +82 54 279 2870