Aaeon IoT gateway improves machine learning

  • November 23, 2022
  • Steve Rogerson

Taiwanese company Aaeon has introduced an IoT gateway to improve machine-learning applications.

The SRG-IMX8P is powered by NXP’s i.MX8M Plus quad-core Arm Cortex-A53 and Cortex-M7 processors. It harnesses the benefits of an integrated neural processing unit capable of producing 2.3Tops.

Further, the inclusion of the Cortex-M7 processor offers real-time calculation functionality to enhance machine learning, reducing visual inferencing time for object detection to just 13ms.

Onboard DDR4L system memory is up to 8Gbyte, plus it has 16Gbyte eMMC storage and a Micro SD slot for storage expansion.

Connectivity includes two USB 3.0, two RJ-45 GbE, and two Can-FD ports.

These produce a configuration for edge gateways in Industry 4.0 settings such as vehicle assembly and factory automation applications.

The USB 3.0 and RJ-45 LAN ports support more sophisticated camera and sensor peripherals, while the Can-FD slots let users transmit and receive sensor data via the IO directly, negating the need for USB-to-Can bus conversion.

To ensure reliable operation across vertical markets, it has an operating temperature of -20 +70˚C, while boasting a power input range of 9 to 36V. The temperature range makes it deployable in any industrial setting, and its power input range opens the door to in-vehicle applications by allowing for stable operation in environments with fluctuating power supply sources.

The gateway has already entered mass production, and is available for purchase, with pricing dependent on SKU.

Established in 1992, Aaeon is a designer and manufacturer of industrial IoT and AI edge products. It provides computing platforms including industrial motherboards, rugged tablets, embedded AI edge systems, uCPE network appliances, and LoRaWan and WWAN products. OEM and ODM services are available worldwide. The company works with cities and governments to develop and deploy smart city ecosystems, offering individual platforms and end-to-end systems.