Powerful ultra-compact AI inference systems with low power computing

25-11-2021 | Advantech | Subs & Systems

Advantech Co., Ltd has launched the AIR-020 series, an ultra-compact series of edge AI systems powered by the NVIDIA Jetson family. Offering a small footprint of only 139mm x 110mm x 44.5mm (5.47" x 4.33" x 1.75") dimensions, the series provides low power computing with powerful AI inferencing capabilities using NVIDIA Jetson Nano, TX2 NX, and Xavier NX SoM. The series is ideal for edge AI applications, including traffic monitoring, AGV/AMR, defect inspection, people counting, medical imaging and more.

It is powered by the NVIDIA Xavier NX SoM, AIR-020T by Jetson TX2 NX, and AIR-020N by Jetson Nano to satisfy diverse AI application and performance demands. The three offerings provide an ultra-compact design with just 139mm x 110mm x 44.5mm (5.47" x 4.33" x 1.75") dimensions, making them simple to integrate into the applications. AIR-020X supports up to 21TOPS and 1058fps aimed at high-resolution imaging processes. AIR-020T and AIR-020N support 1.33/0.5 TFLOPS and up to 109/48fps. The series supports wide-range 12-24VDC input, -10C to 55C operating range, and vibration and humidity resistance.

AI projects are employed in a wide variety of applications needing variable peripheral connections. The series is provided with extensive I/O ports incorporating 2 x USB 3.2 Type A, USB 3.2 Type C, 1 or 2 x GbE, 1 or 2 x RS-232/422/485, 8-bit DIO, and CANBus for data acquisition and communication. The ultra-compact AI boxes are pre-installed with 4GB/8GB LPDDR4 and 16GB eMMC 5.1. Furthermore, it offers 128GB of M.2 storage as the data storage default setting for AI usage.

The series is bundled with the Ubuntu 18.04 LTS OS environment, Advantech Edge AI Suite and JetPack SDK 4.5.1. Developers can simply develop AI models and deploy applications. For data security, it adopts TPM 2.0 trusted platform module and secure boot, which opens the door for compute-intensive industrial applications and greater system security on the edge.