direct go
menu direct go
contents direct go
sub menu direct go

ETRIElectronics and Telecommunications Research Institute

Organization

Introduction of Research

Autonomous Manufacturing Research Division

  • Behavioral Intelligence for Autonomous Manufacturing Technology
    The Behavioral Intelligence for Autonomous Manufacturing Research Section develops Physical AI technologies that enable manufacturing robots and equipment to perform tasks precisely and safely in real-world environments. We build high-quality training datasets using data-centric AI and address real-world data limitations through automated virtual environment generation, physics-based simulation, and synthetic data generation. Sim2Real and Real2Sim help reduce the gap between real and virtual environments. We also investigate fine-tuning and field adaptation of Physical AI models trained on real-world and simulated data, alongside error detection and safety management. Through these efforts, we aim to realize intelligent autonomous manufacturing systems that integrate high-quality dataset development, model training, and safe deployment in manufacturing environments.

    Key Research Areas
    - (Data-Centric AI) High-quality training dataset development from manufacturing data through automated metadata generation and labeling, data quality management, and data augmentation.
    - (Virtual Manufacturing) Automated construction of virtual manufacturing environments, physics-based simulation, synthetic data generation, and Sim2Real/Real2Sim integration to produce high-quality training data.
    - (Embodied AI) Model fine-tuning, error detection in Physical AI systems, safety management, and field adaptation for precision manipulation and safe control of manufacturing robots and equipment.

    Behavioral Intelligence for Autonomous Manufacturing Technology Image <>
  • Autonomous Collaborative Operation Intelligence Technology
    The Autonomous Collaborative Operation Intelligence Research Section develops autonomous collaborative operation technologies that enable heterogeneous Physical AI systems in manufacturing environments to understand situations on their own and work together to complete tasks. We perceive manufacturing environments and resource states to analyze and resolve collaborative situations, and perform collaborative planning, decision-making, and orchestration among heterogeneous Physical AI systems. We also research embodied behavior intelligence that generalizes to unseen objects and environmental changes, and develop robot manipulation technologies based on demonstration-replay and data-efficient post-training. Underpinned by on-device and edge AI-based distributed collaboration, autonomous learning, and trust and safety technologies, our goal is to build an autonomous collaborative operation framework applicable to real-world manufacturing sites.


    - Manufacturing environment and resource state perception, and collaborative situation analysis and resolution
    - Collaborative planning, decision-making, and orchestration among heterogeneous Physical AI systems
    - Generalization of embodied behavior intelligence to unseen objects and environmental changes
    - Manufacturing robot manipulation based on demonstration-replay and data-efficient post-training
    - Edge AI-based distributed collaboration, autonomous learning, and trust and safety

    Autonomous Collaborative Operation Intelligence Technology Image <>
TOP