From Advantech for the industry
Compact Edge AI system for fast AI introduction
Advantech's compact AIR-120 edge AI inference system enables rapid implementation of edge AI applications. It is powered by Intel Atom-x7433RE (Amston Lake), Core-i3-N305 (Alder Lake-N) and N97 (Alder Lake-N) processors, as well as the Hailo-8 M.2 AI accelerator module.
The system configuration offers scalable performance with up to eight cores and up to 26 TOPS of AI computing power. It enables high-throughput, low-latency inference with a favourable balance of performance, energy efficiency and cost – suitable for vision AI applications, for example.
Designed for industrial environments, the AIR-120 is available now and supports a wide operating temperature range from -20 to 60 °C and complies with the EMC standards IEC 61000-6-4 and IEC 61000-6-2 for heavy industry. To simplify the implementation of edge AI applications, the AIR-120 comes with the Advantech Edge AI SDK, Advantech EdgeBMC (Edge Baseboard Management Controller) and the Hailo AI Software Suite. The tools provide benchmarking utilities, runtime SDKs, and remote monitoring and management capabilities, simplifying model development, conversion, optimisation, and deployment. Target applications include public safety monitoring, AI-assisted security inspections, and employee safety.
Important features of the AIR-120
The AIR-120 combines low energy consumption, dedicated AI acceleration and industrial robustness. This makes it a cost-effective, easy-to-implement edge AI inference system that integrates seamlessly into existing infrastructure. Its compact housing (176 x 112.35 x 66.4 mm) accommodates AI computing functions, remote management functions and more than twelve I/O options to support cameras, device control and communication interfaces. The I/O interfaces include 1 x 2.5 GbE, 1 x 1 GbE, 2 x COM, 2 x USB 3.2, 2 x USB 2.0, optional dual CAN bus and 3 x M.2 expansion slots. One of the Ethernet ports supports EdgeBMC for out-of-band remote management of the system. The internal M.2 slots enable easy integration of the Hailo-8 AI acceleration module as well as storage devices, Wi-Fi and 4G LTE modules. A special thermal solution ensures stable performance of the Hailo-8 under high AI workloads, such as YOLOv3-based object detection, preventing overheating and ensuring reliable operation.
Seamless AI deployment with software toolkit integration
The AIR-120 integrates the Hailo-8 AI acceleration module with a comprehensive suite of development tools to enable rapid AI deployment. The Advantech Edge AI SDK is fully integrated with the HailoRT runtime SDK, enabling quick setup and optimised development processes. For AI model conversion, the Hailo AI Software Suite offers:
• Model Zoo: A collection of ready-to-use models;
• Dataflow Compiler: For efficient cross-platform model conversion;
• TAPPAS Toolkit: A library of pre-trained models for applications such as multi-stream object detection, distance segmentation, classification and pose estimation.
For further development support, the Edge AI SDK offers benchmarking tools and AI reference examples for evaluating hardware performance. In addition, integrated support for EdgeBMC enables out-of-band remote management, which simplifies monitoring and maintenance for scalable edge AI deployment.
