Taiwanese IoT firm Advantech has launched a compact multimodal vision AI edge inference system to accelerate AI adoption in smart manufacturing, robotics, automated optical inspection, smart transportation and smart retail applications.

The Air-055 is powered by Qualcomm Dragonwing IQ-9075M processors.

As edge AI applications evolve from single-purpose image recognition to multimodal AI, enterprises increasingly need compact platforms that can process cameras, sensors, network data and control signals in real time. The Air-055 addresses this by combining 100Tops AI performance, extended-distance camera connectivity, more than 20 IO interfaces and expansion slots, and integrated software tools in a compact box.

By integrating Advantech WEDA (Wise-Edge Developer Architecture), edge AI SDK, BSP launcher and Qualcomm AI hub resources, it can lower application development barriers, shortening the path from proof of concept to mass deployment.

It has 36Gbyte LPDDR5 memory, improving memory use by 1.5 to 4.5 times across AI models. With 60W power consumption, it accelerates generative AI, computer vision and multi-model inference at the edge. For vision AI applications, the system also supports USB and PoE industrial cameras and can be integrated with GMSL modules for longer-distance or harsh-environment deployments.

This makes it suitable for AOI, AMR perception, robotic vision and smart manufacturing applications requiring low-latency image capture and real-time AI processing. IO interfaces and expansion options, including USB, LAN, COM, DIO, Can, display outputs, M.2 NVMe, Wifi 6E, Bluetooth, 5G LTE and PoE expansion. Its 190 by 174 by 57.8mm fanless design supports -20 to +55˚C operation, making it suitable for round-the-clock industrial deployment in space-constrained environments.

To simplify AI development and large-scale deployment, it integrates Advantech’s software tools, including BSP launcher, edge AI SDK and WEDA. The BSP launcher supports remote BSP updates to reduce onsite maintenance effort. The edge AI SDK provides pre-validated AI frameworks and runtime environments, while supporting more than 300 optimised models from Qualcomm AI Hub to accelerate model selection, integration and validation. With WEDA, it enables remote deployment, OTA updates and device management, helping enterprises reduce the complexity of managing edge AI systems at scale.

With more than 9000 employees worldwide and a 40-year track record, Advantech (www.advantech.com) specialises in embedded AIoT.