Solution

Edge Computing

AI inference, protocol gateways and pre-processing where the data is created — not in a data centre three hundred kilometres away.

Industry pain points

What typically goes wrong in this industry — and what it costs you.

1
Cloud round-trip latency
Sending every frame to the cloud adds 200–600 ms and breaks real-time control loops.
2
Bandwidth cost
Raw video and sensor streams saturate uplinks; metered 4G/5G turns into a monthly line item.
3
Harsh edge sites
Street cabinets and substations run −20 °C to 60 °C on unstable power.
4
Fragmented protocols
Modbus, MQTT, OPC UA, CAN — one gateway is expected to speak all of them.

Recommended models

Platforms we ship for this scenario, with the I/O and thermal headroom it needs.

G6-F1-RK3588
Edge AI Box · RK3588 · 6–32 TOPS
6–32 TOPS NPU for on-device vision inference
View model ›
K8-F24
IIoT Gateway · Pentium · 4 × COM
IIoT gateway, 4 × COM for Modbus and serial telemetry
View model ›
G4C-RK3566
ARM Controller · RK3566 · RS485 / GPIO
ARM controller with RS485 / GPIO for field I/O
View model ›
RK3588 Board
Nano-ITX Motherboard · RK3588
Nano-ITX board for custom enclosures and OEM builds
View model ›

Application cases

On-device AI inference running on a MINI X edge AI mini PC
Smart city traffic analytics · 180 intersections
G6-F1-RK3588 runs plate and flow detection locally; backhaul traffic dropped 92 % versus cloud processing.
Fanless heatsink and thermal module inside a MINI X mini PC
Substation monitoring · 45 sites
K8-F24 gateways aggregate Modbus and IEC 104 telemetry across −20 °C to 60 °C sites.

Need a configuration for your project?

Tell us the environment, the I/O list and the volume — a project engineer replies within one business day.

Request a Quote →
WhatsApp