From Robotic Automation to Smart City, Advantech is brining Real-Time AI to the Edge

When entering a warehouse of an e-commerce company, it is normal to see crowds moving and goods going between the rows of shelves. In this case, the employees were relying on the use of several automated guided vehicles (AGV) to do the work for them. A glimpse, there weren’t tracks on the ground, nor QR codes or bar codes nearby to guide the route. Each AGV knew how to get to the shelves at the nearest isle because each AGV can detect if a barrier or person is near and can quickly detour or stop itself.

Making Factory AGVs More Efficient and Safer with AI-based Navigation

“The secret lies in the edge AI solutions,” said Stephen Huang, associate vice president of the Advantech Embedded-IoT group. Each AGV integrates VEGA-330, Advantech’s edge AI acceleration miniPCIe module. It can recognize and process images captured from the camera, and AI navigation allows it to decide the best route in real-time, boosting each AGV’s runtime efficiency. As cameras are the eyes of AGVs, Advantech edge AI acceleration modules act as the engine of visual analytics and optimize image signal processing and inference on the device to help each AGV execute better route-planning and auto-stop before a collision occurs. They can also self-learn in real-time. For e-commerce, logistics, and the warehousing industry, AGVs with AI-based navigation improve efficiency by up to 20%. It also enhances the safety of operation as people or other vehicles are detected and possible collisions can be avoided.

Stephen pointed out that Advantech is helping customers to transform their traditional AGVs to AI-based AGVs with better space utilization and optimized battery life while using fewer development resources.

Improving Metropolis Traffic with AI-powered Street Lighting

Next, let’s change the scene from the warehouse to a corner of the city. Thousands of street lights illuminate the city helping drivers and pedestrians find their way home safely. These street lights not only light up the city. They can also provide city services including air quality, humidity/temperature monitoring, and traffic and parking management. Imagine if people could know the information of the parking spaces on the roadside from their mobile phones. They could drive directly to an available parking space without searching. To make it possible, we need to leverage the power of edge AI. 

Stephen explained that the customer already implemented Advantech’s edge computing platform. Sensors and surveillance cameras were installed in the street lamps to collect 24/7 information such as temperature, humidity, air quality, and roadside conditions. Information was then uploaded to the cloud for operational analysis. Now, Advantech edge AI acceleration modules (VEGA-320), which are embedded into the cameras can analyze video data from multiple streams and handle real-time tasks such as locating open parking spaces in urban areas. The customer estimated that the utilization efficiency of roadside parking was increased by at least 10-15%. In addition to roadside parking space management, other metropolitan-wide services including real-time traffic flow analysis, real-time traffic accident reporting, and traffic law enforcement could be created.

Although the scenarios in the two cases above are quite different, they both use the VEGA-300 edge AI acceleration module of Advantech for its small size, low power consumption, and scalable architecture. Powered by Intel Movidius X VPU (Vision Processing Unit), Advantech VEGA-300 series has a small footprint in only 8.1mm x 8.1mm, and less than 3W power consumption, making it the best fit for edge AI applications. In addition, since the number of Intel Movidius X VPU units onboard can be expanded, system integrators could either add or reduce VPUs in accordance with different computing needs. Therefore, both AGV and smart street lighting projects could easily adopt VEGA-300 without significant changes in the original design. 

The movement of real-time computing, data security, bandwidth capacity, and energy-saving, has allowed edge AI to become a powerful new technology. Advantech has launched multiple solutions to fulfill such market needs. In addition to edge AI acceleration modules, there are requirements for embedded platforms and edge computing systems with AI capabilities. For example, to achieve the needs of vision inspection in a factory production line, the performance of the edge AI system would need a server-grade Xeon processor to support its real-time operation. An additional high power GPU card acceleration engine would also be necessary. Advantech’s AIR-300 is the ideal solution for this project because it is equipped with an Intel Xeon E3 processor and EC memory support. One 260W NVIDIA high-performance GPU card can also be installed into AIR-300. This can ensure high-density image processing and inferencing in real time and help keep production highly efficient. Furthermore, Advantech also provides Edge AI Suite software that works with the Intel OpenVINO toolkit. System integrators can easily install Edge AI Suite in just a few clicks and experience the power of Advantech’s AI solutions. It incorporates ready-to-use programs such as facial recognition and human pose detection. It also provides CPU, GPU, and VPU usage information.