Advantech, Basler and Canonical jointly explore three key factors in the development of robotics

Advantech's Embedded Single Board Computer Robotics Online Forum Concludes Successfully

Keywords: Advantech embedded single board computers, robots, machine vision, AI, AI computing, logistics and warehousing, unmanned factories, automated production, AGV, AMR, AOI, optical inspection, visual navigation

May 2022, Shenzhen, China - On May 11, Advantech, together with its partners Basler and Canonical, successfully held an online forum on the theme of "Three Key Elements in the Development of Robots: Vision, Nerves and Brain", discussing with everyone the three key elements in the development of robots. This article will take you through the highlights of the meeting!

According to the Omdia/HIS 2021 market research report, the total market value of mobile robots (AGV/AMR) will reach US$6 billion by 2024. Among them, mobile robots using traditional magnetic stripe sensing and QR code technology will each account for about 12% of the overall market value, and the overall market value of robots using simultaneous localization and mapping technology (SLAM) will reach about 50%.

For this online forum, Advantech specially invited experts from Basler, a camera manufacturer that plays a key role in the development of robots, and Canonical, a software operating platform manufacturer, to share with everyone the three key aspects of robot application development: "machine vision", "software platform" and "AI computing platform" based on practical applications.

The robot’s eyes: selection and standards of industrial cameras

Li Yujie, business manager of Basler Business Unit, worked back from the visual application end to the required camera technology and specifications, saying "Based on the application requirements of the camera selection factors, we should select the appropriate camera resolution, frame rate (speed), transmission interface, image sensor type, etc. to determine the appropriate camera."

For example, in defect detection AOI systems, since the camera needs to capture tiny defects at high speed, it is recommended to use CoaxPress interface, high resolution, high frame rate, high bandwidth interface and anti-interference wire. For shooting at low speed or static state on the robot arm, GBE interface, lower frame rate camera can be selected, and high-interference wire must be used due to the twisting of the robot arm. The resolution is determined by the size of the object being measured.

We see that AMR, mobile robots, automatic assisted driving and other complex usage environments require depth perception and 3D vision. In this regard, Mr. Li Yujie shared the following on "How to choose 3D imaging technology suitable for the industrial field":

1. Time of Fly (ToF): Relies on the time of light flight to measure distance. Suitable for applications in robotic arms, AMR, and AGV logistics.

   
   

2. Structured Light: Deformation of the projected pattern. Suitable for 3D defect inspection (AOI) and 3D scanning.

3. Active/Passive Stereo Vision: Corresponding points from two camera images. Suitable for robotic arms and service robots.

The nerves of robots: How Ubuntu supports developers and ROS

Regarding embedded operating platforms and software, how to achieve real-time security updates is a major challenge for robots deployed in large numbers. Canonical networking solution architect Peng Riting said that Ubuntu provides long-term maintenance and security operating systems. To ensure security, there are public and objective security testing indicators (CVE), and security updates are released regularly every three weeks, with a maintenance cycle of up to 10 years.

In addition, the ROS/ROS2 used in robot applications also provides an adapted Ubuntu LTS version for long-term update and maintenance. Based on Ubuntu OS, the AI ​​robot platform can provide customers with a variety of AI SDKs and a complete software development ecosystem during the development phase. During the training phase, a large amount of data is required. Ubuntu OS has close cooperation with various cloud service providers such as Azure AWS and Oracle, so users can easily train AI models. On the ground side, there are many device options. The development experience on Ubuntu will be very complete and convenient.

Advantech has currently widely adopted the Ubuntu operating system for verification on hardware platforms, and has a cooperation agreement with Canonical to provide Ubuntu Certified certification.

The robot’s brain: Advantech’s embedded high-efficiency AI computing platform and services

Wang Shengwen, senior product manager of Advantech's Embedded IoT Platform Business Group, shared that from the application perspective, the current mobile robot development platform can be roughly divided into two types: 1. Built-in AI performance: suitable for robotic arms or AMRs, and 2. Expandable AI acceleration cards, which can expand xPU, FPGA, etc. Based on the characteristics required for developing a robot platform, we develop Advantech's embedded AI platform based on the AI ​​computing capabilities required by different applications, using space and environment.

• Miniaturized robotic arms, commercial drones, etc. require miniaturization and lightweight features. We provide the first-generation Pico-ITX AI computing platform MIO-2375, which is only the size of a palm.

• AGV, AMR, delivery robots and other applications in highly complex indoor or outdoor environments do not require high size, but need to process a large number of data streams and sensors. We provide the 3.5” highly integrated embedded single board MIO-5373 starting from Intel 11th generation to meet various needs.

• In the future, specialized fields will require more efficient AI computing, which will require the expansion of specialized accelerators such as independent GPUs, xPU accelerators, ASICs, FPGAs, DSPs, and other accelerators. We are also developing the 12th generation Intel platform, such as MIO-4370, which provides PCIe Gen.5 and Thunderbolt expansion capabilities.

• Advantech has launched the EPC-C301 palm-sized fanless embedded industrial computer, which is based on Intel's 8th generation low-power ultrabook CPU and integrates a rich set of I/O interfaces to fully meet the application requirements of AGV and AMR.

Compared with server or cloud AI computing, the application of embedded AI is more important in terms of the balance between performance and power consumption, size and cost, timeliness and benefits. Advantech is committed to providing a platform suitable for embedded AI applications, from native AI computing performance to scalable AI computing solutions, and with Ubuntu and Yocto OS , it greatly reduces the difficulties of customers in early development. Advantech will also continue to work with partners to provide customers with more and better AI and robotics solutions and technical support.