Recently, interest in the industrial environment has focused on automation and artificial intelligence (AI). AI, which is expected to reduce costs and replace human labor, is being commercialized in various fields through technological advancements. Developers need efficient edge computing systems, and there is a growing demand for systems that can be updated quickly without complex processes.

At the 'Edge AI-Based Digital Innovation Strategy' webinar held by Advantech on the 12th, industry experts participated and presented edge AI trends, market insights, and related solutions.

AI experts, including NVIDIA Executive Director Ham Heung-yong, Advantech IIoT Headquarters Director Kim Kyung-yeon, and Nota Director Yang Seok-yeol, participated to analyze market needs and provide relevant insights.

■ Demand for real-time monitoring surges, leading to a rise in edge AI.
Edge AI is a technology that processes data at the network edge near the user, unlike conventional cloud-based AI. As demand for real-time monitoring of IoT devices rapidly increases due to the spread of Industry 4.0, edge AI applications are expected to increase significantly.

In a presentation, NVIDIA's Executive Vice President Ham Heung-yong said that NVIDIA has prepared an edge AI platform with an ultra-small and ultra-low power form factor, so that it can be applied to various industries.

The connection between smart factories and edge AI was introduced by showing how IoT is being used to connect, collect, and analyze production lines and data, evolving from mass production to customized small-batch production at minimal cost and time.

In this process Edge AI enables data collection and inferencing through application services, and automation is implemented using digital twins and techniques, enabling real-time processing, intelligent tasks, and improved data security and efficiency.

These technological advancements are driving the true meaning of smart factories, and NVIDIA's Jetson hardware platform provides a diverse lineup of end-to-end SDKs (Software Development Kits) to support them, he said.

Edge AI: Strengths in Minimizing Bandwidth, Enhancing Security, and Reducing Construction Costs
Advantech Industrial IoT Headquarters Manager Kim Kyeong-yeon introduced the MIC series, a product that supports process acceleration from AI development to deployment, and explained application cases.

Edge computing-based AI is being utilized to simultaneously perform input, distribution, and processing at each edge for applications that were previously centralized on servers. Accordingly, it is expected that the smart city and industrial internet sectors will lead the growth.

Edge has four main strengths. First, it minimizes network bandwidth, thereby facilitating the application of on-time applications by simultaneously processing network overload and bandwidth burden caused by transmitting and receiving large amounts of image and video files at the edge.

Second, it offers advantages in cyberattacks and data security. Third, it minimizes infrastructure construction costs. Finally, it facilitates error analysis, enabling immediate detection of issues at each edge. Maintenance is also simple.

Advantech's AI platform, based on NVIDIA Jetson, launched three new models in June: the Orin NX, Nano, and AGX. The most recent Orin series boasts approximately nine times the performance of the earlier Nano, achieving performance comparable to entry-level AI server GPUs. It is expected that the specification structure will be widely applied to applications that use x86 Windows-based systems.

The advantages of the Advantech brand include: △providing stable products as a global leader in industrial computers and designs suitable for edge AI; △possessing references that can be used in various applications; △configuring all AI architectures with only Advantech products and meeting diverse market needs; △providing technical support to customers by supporting NVIDIA's BSP updates; and △possessing a global AI partner network for joint development in the ecosystem.

Netpresso automatically creates hardware-optimized AI models.
NetsPresso is a platform that automatically improves hardware-optimized models. It provides technology that automatically generates an AI model optimized for the hardware by inputting data and target performance.

In the presentation by Nota's Director Yang Seok-yeol, Nota's Netpresso solution and application cases were introduced. The Netpresso platform is explained to be composed of three modules, with automated machine learning, deep learning, and engineering resources.

The first module, 'Model Searcher', generates an optimized AI model based on data through AutoML function, and the second module, 'Model Compressor', compresses the previously generated model more deeply to make it lighter.The last 'Model Launcher' provides the function to load the optimized model onto actual hardware.

Customers running cloud-based AI systems can leverage Netpresso to optimize and streamline cloud costs, while customers deploying AI models on edge hardware can also leverage model optimization to reduce hardware costs and improve competitiveness.

Applications of these solutions include driver monitoring systems used in the mobility sector, which can monitor drivers for drowsy driving, smoking, and neglectful driving behavior, such as operating a mobile phone. This is an area where collaborations with vendors, such as black box manufacturers and automakers, are ongoing.

Additionally, video analytics solutions based on road cameras are being applied in the ITS (Intelligent Transportation System) field, optimizing signal control through traffic volume and congestion analysis at smart intersections and crosswalks. Video analytics solutions are also widely used for parking space management through parking lot monitoring.

Director Yang Seok-yeol stated, “These solutions are being applied both domestically and internationally, and in the future, we plan to expand our business in the CCTV and healthcare fields by expanding the data scope to voice and NLP (natural language processing) through AI lightweighting and optimization.”

Meanwhile, the webinar held that day can be replayed on demand on Advantech's video platform, A-Connect.