Overview
Overview
Industrial Edge Controller is a ruggedised, fanless industrial PC that runs Docker / Kubernetes containers next to the machine, performs AI inference on the production line, and closes a real-time control loop with sub-millisecond determinism. One box replaces the traditional "PLC + edge gateway + vision PC + protocol converter" quadrumvirate that clutters most modern cells.
Built for plants modernising beyond classic PLCs — vision-guided pick & place, predictive maintenance with ML, real-time energy optimisation, OPC UA aggregation, and tight integration with the cloud GitOps pipeline that engineering teams already use. Linux-based, container-native, and managed from a central fleet console.
- Container-native — Docker / Podman / k3s lightweight Kubernetes
- AI inference at the edge — ONNX / TensorRT / OpenVINO on optional GPU or NPU
- Real-time co-located — Xenomai / PREEMPT_RT kernel for hard real-time loops
- GitOps managed — Argo CD / Flux push updates to a fleet of hundreds of edges
How It Works
A real-time Linux kernel hosts a normal Docker / k3s runtime alongside a hard-real-time scheduler. Field protocols and control loops live in privileged real-time containers; everything else — AI inference, dashboards, telemetry — runs as standard pods. Cloud GitOps deploys new versions; failures auto-rollback.
- Real-time core — isolated CPUs pinned to the RT scheduler with sub-millisecond jitter
- Field I/O — Modbus, Profinet, EtherCAT, OPC UA, GigE Vision direct to hardware
- Inference pipeline — cameras / sensors → preprocess → AI model → control decision
- Container fleet — k3s manages pods, restarts on crash, rolls back on health-check fail
- Cloud orchestration — Argo CD watches the Git repo and reconciles the entire fleet
Features
- Hard real-time — PREEMPT_RT kernel, isolated cores, < 100 µs typical jitter
- GPU / NPU options — NVIDIA Jetson, Intel Arc / Movidius, Hailo-8, Coral Edge TPU
- Cyber-resilient — TPM 2.0, secure boot, encrypted disk, signed container images
- OTA / GitOps — declarative updates, canary rollouts, automatic rollback
- Standards-based — OPC UA Server / Client, MQTT 5, REST, Sparkplug B
- Observability — built-in Prometheus exporters, OpenTelemetry traces, Loki logs
- Vision-ready — GigE Vision, USB3 Vision, MIPI camera inputs with hardware sync
- Fanless design — passive cooling for dusty / vibrating environments, 24/7 duty
Use Cases
- Vision-based quality inspection — YOLO / Mask-RCNN on the line at 30 FPS, reject lane control
- Predictive maintenance — vibration / current pattern recognition, raise ticket before failure
- Energy optimisation — closed-loop control of HVAC, compressed air, chillers using ML setpoints
- Robot guidance — 3D / 2D vision feeding robot trajectories in real time
- Process aggregation — one edge controller per cell talking OPC UA up to plant SCADA
- Edge digital twin — high-frequency mirror of the physical asset for what-if simulation
Specifications
- CPU: Intel Xeon-D 8-core / Intel Atom x6000 / NVIDIA Jetson AGX Orin options
- Memory: 8 / 16 / 32 / 64 GB DDR4 ECC, 256 GB to 2 TB NVMe
- GPU / NPU: optional NVIDIA RTX A2000, Hailo-8L 26 TOPS, Coral 4 TOPS
- OS: Ubuntu Core / RHEL / Yocto with PREEMPT_RT, k3s 1.28, Docker 24
- Ports: 2–4× 2.5G Ethernet, 4× USB 3, DP/HDMI, RS-485, CAN, 4-DI/4-DO
- Power: 24 VDC ±20 %, 25–80 W typical, optional redundant input
- Environment: −25 to +60 °C, IP30, fanless, 5 g vibration, M12 / M8 connector options
- Security: TPM 2.0, secure boot, FIPS 140-2 crypto, signed firmware, role-based SSH
Product information
| Product name | Industrial Edge Controller |
|---|---|
| Category | Factory Automation |
| Type | Hardware |
| Procurement | Contact / Negotiable |
Customisation options
- CPU class — Intel Atom (low-power), Xeon-D (server-grade), or NVIDIA Jetson Orin (GPU-AI)
- AI accelerator — none, Coral Edge TPU 4 TOPS, Hailo-8L 26 TOPS, NVIDIA RTX A2000, or Jetson Orin 275 TOPS
- Storage — 256 GB / 512 GB / 1 TB / 2 TB NVMe, optional dual-disk RAID-1 for data integrity
- Real-time guarantee — standard Linux, PREEMPT_RT < 100 µs jitter, or Xenomai < 20 µs for hard motion control
- Connectivity — add Profinet master, EtherCAT master, GigE Vision quad, 5G modem, Wi-Fi 6E
- Cluster mode — single node, HA pair (active/standby), or k3s edge fleet with central GitOps
- Cyber stack — standard, FIPS 140-2 crypto, IEC 62443 SL-3 hardening pack
- Mounting — DIN-rail, VESA arm, wall-mount, or 1U rack tray; M12 connector option for vibration
Ordering & lead time
Standard lead time depends on current stock. For volume orders, please contact our team for a structured quote including BOQ, freight, installation and warranty terms.
Support
Engineering support, installation guidance and warranty claims are handled by our customer success team. Reach us at support@sgtsystems.com or via the channels listed on our Contact page.