AI Model · AI Agent
We study Vision AI, Domain AI, and AI Agent for on-site judgment and planning according to industrial purposes.
CORE TECHNOLOGY PLATFORM
Industrial Physical AI
PPC — Perception · Planning · Control
We design AI models according to field purposes, implement 2D/3D Industrial Perception and NPU Edge AI into products and facilities, and connect the execution that has passed Validation/Safety with the industrial control layer.
AI RESEARCH → ENGINEERING → PHYSICAL SYSTEM
We research AI and connect it to field computing environments and actual facilities.
Industrial Physical AI of ILLUVATION takes the AI Model·AI Agent research as its starting point and connects to recognition, judgment, and safe execution of actual industrial systems through engineering and edge placement tailored to the purpose, data, and computational constraints.
Model Engineering · NPU Edge
Design and redesign the model structure by considering the problem, data, target performance, and computational constraints and deploy it in the NPU ·Embedded environment.
Perception · Planning · Validation/Safety · Control
The sensor, AI, and control layers are connected to actual facilities and the execution results are reflected back as feedback.
Capability Chain AI Research → AI Model Engineering / Agentic AI → NPU Edge · Embedded AI → Industrial Physical AI
REFERENCE ARCHITECTURE
Connects everything from research to physical systems into one technological system.
AI Model·AI Agent Based on research and model engineering, it connects Perception, Planning, Validation/Safety, Control, Physical System, and Feedback into one closed loop.
COLLABORATION
Scope of technology/R&D/SI cooperation
AI Model Engineering
Model structure design and redesign tailored to Domain AI and purpose, data, and computing environment.
2D·3D Industrial Perception
Camera, RGB-D, Depth, state, object, and space recognition using 3D and field sensors.
NPU · Embedded AI
Placement and optimization of Edge AI for NPU level computing environment inside products and equipment.
MCU · PLC · PPC Integration
Validation/Safety, control/system integration that connects MCU/PLC/Controller with actual equipment.
PPC ARCHITECTURE
From Perception to Feedback
It connects on-site awareness, operation planning, safety verification, actual facility execution, and feedback in a closed loop.
Awareness of site conditions
Objects, status, and changes are recognized based on camera, RGB-D, 3D sensor, and field sensor data.
→
Rule-Set · Agentic Planning
Select a predefined Rule-Set or, in complex situations, use JENNA Agentic AI to create and adjust an execution candidate Rule-Set.
→
Validation/Safety · Deterministic execution
After verifying the execution candidates with safety conditions, they are executed in the industrial control layer such as Edge Controller, MCU, and PLC, and the results are fed back.
CORE AI MODEL ASSETS
Own AI model assets
Domain AI, Industrial Perception AI, and Agentic Planning AI are selected and combined according to industrial problems.
Livestock specialized AI
High-Precision Weight Inference AIRGB-D Object weight inference AI using 3D data
Real-Time Weight Inference AIAI that uses continuous RGB-D data to infer the weight of multiple pigs in real time
Farrowing Monitoring AIApplication of AI · PigTocTok to monitor the calving status of one sow
universal visual perception
SwiftYOLOGeneral-Purpose Object Recognition AI developed by ILLUVATION
JENNA Agentic AI
GPU-based Agentic PlanningOn-Premise Agentic AI for Local-First Engineering. Supports Rule-Set creation, modification, and optimization in complex situations.
PLANNING
Deterministic Rule-Set · JENNA Agentic AI
Planning optionally configures Predefined Rule-Set and GPU-based JENNA Agentic AI depending on site complexity.
Predefined Deterministic Rule-Set
In sites with clear operating and safety conditions, an action plan is constructed using a pre-verified Rule-Set.
- Explainable operating conditions
- Verifiable control logic
- Stable operation based on fixed rules
JENNA Agentic AI
We support Rule-Set creation, modification, and optimization in sites where operation plan adjustments are required according to multiple conditions and situational changes.
- Context-based planning support
- Output results in Rule-Set format
- Apply validation/safety before execution
Agentic Planning · Deterministic Control
The Rule-Set created by JENNA is verified by the Validation/Safety layer and is executed deterministically in the industrial control layer.
LOCAL-FIRST ENGINEERING
On-Device · On-Premise · On-Site Control
NPU Edge Perception
Necessary data collection and on-site inference are performed in a NPU-class computing environment close to cameras and sensors.
Planning / Local AI Runtime
Configure planning functions such as JENNA and Rule-Set management in the field local environment.
Deterministic Control
Verified Rule-Sets are executed on PLC, MCU, Edge Controller, and field control infrastructure, and safety interlock and emergency stop systems are maintained.
R&D DIRECTION · DIGITAL TWIN
Industrial Physical AI to DTBS
Currently, through Digital Twin research, we are researching ways to express objects, spaces, facilities, and process states and update them with feedback. DTBS — Digital Twin Based System is an advanced target system.
Industrial Physical AI
Perception → Planning → Validation/Safety → Control → Physical System → Feedback.
Digital Twin
Research on expressing objects, spaces, facilities, and process states in Digital Twin and updating them with on-site feedback.
DTBS
Closed-loop goal system linking Digital Twin status to Planning and Control.
PROVEN IN THE FIELD
Algae removal equipment Industrial Physical AI integrated control
In 2026, we have built a closed-loop Industrial Physical AI system that connects RGB-D·RADAR·CCTV-based recognition, on-site AI Agent, MCU safety/control layer, PLC interlock, and actual facility feedback at environmental infrastructure sites in 2026.
PERCEPTIONRGB-D · RADAR · CCTV
AI AGENTSAlgae Agent · Hoist Agent
SAFETY & CONTROLMCU · PLC Interlock
PHYSICAL SYSTEMGate · Ejector Pump · Plasma System · Hoist
FEEDBACKSensor · Equipment State · Event · CCTV
The judgment of AI Agent is executed as the final equipment output after verification of MCU safety/priority control and PLC interlock.
COLLABORATION
Industrial Physical AI Joint implementation
We discuss AI joint research, 2D/3D Perception/Sensor/Edge combination, MCU/PLC/equipment integration, SI construction, and the scope of joint R&D.
Technology/R&D/SI cooperation consultation