Controller Integration & Edge Computing

Concept & Purpose

Standard industrial controllers (PLCs) handle millisecond-level physical tasks (driving motors, reading sensors) but lack the compute capacity, file-system drivers, and neural inference engines to execute AI logic.

YAIFA solves this by introducing a demarcation architecture:

• The Brain (YAIFA): Executes cognitive BDI reasoning and model scoring on an industrial Edge PC close to the machinery.
• The Hand (PLC): Executes safety loops and direct motor drivers, writing raw data to YAIFA and reading setpoints.
• OT Integration: Communication occurs locally inside the OT network namespace using OPC-UA or MQTT, keeping data off public cloud connections.

YAIFA creates and versions the "brain" of the agent (BDI logic, ML models, decision and communication code). The PLC or edge gateway serves sensors and actuators (I/O, fieldbus, OPC-UA server on the machine). The brain decides and communicates; the PLC executes physical action. This architecture separates responsibilities and clearly defines the interface (typically OPC-UA, MQTT, REST).

Practical Value

Edge-level deployment guarantees reliability and safety:

• Siemens & Bosch Rexroth Support: Compiled container packages (Docker images) are optimized for industrial edge gateways, such as the Siemens SIMATIC IOT2050 and the Bosch Rexroth ctrlX CORE.
• Legacy Integration: For older machinery lacking native OPC-UA servers, YAIFA acts as an intermediate gateway in the local network, translating fieldbus signals into normalized JSON variables.
• Safe-Failover (Degraded Mode): If communication between YAIFA and the PLC breaks, the PLC reverts to safe hardcoded parameters, and the YAIFA agent logs a critical error, preventing uncoordinated machinery actions.

Technical Realization

Edge deployment coordinates the BDI cycle with local containerized drivers (standard deployment model):

Hardware Level	Software Component	Active Protocol	Task
Edge PC / Gateway	Docker Container (YAIFA Runtime)	OPC-UA Client / MQTT Broker	Reads raw sensor states, updates BDI beliefs, selects optimal setpoints, and publishes targets.
Machine PLC	Local PLC Logic (Structured Text)	OPC-UA Server / Modbus	Pulls setpoints from the Edge broker, verifies mechanical safety interlocks, and drives actuators.

A standard reference Dockerfile makes deployment repeatable:

# Reference Dockerfile for YAIFA Edge Runtime
FROM python:3.13-slim

WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY . .

# Start the generated standalone agent executable
CMD ["python", "agent_main.py"]