ADLink: Move from Embedded to Edge Compute Means Switch from Automation to Autonomy

Article By : Sally Ward-Foxton

ADLink partners with Foxconn on edge compute for robotics and autonomous driving.

At Embedded World Digital 2021, ADLink showcased its edge compute solutions for machine vision, autonomous systems and robotics. The latest hardware includes GPU-based AI acceleration platforms based on Nvidia Jetson and Quadro, new AMD and Intel based AI-on-modules (AIoMs) for applications like ultrasound image processing and video streaming/encoding, and new I-Pi SMARC i.MX-8M+ starter kits.

When EE Times Europe met with ADLink in person at last year’s Embedded World in Nuremberg, the Taiwanese company’s CEO Jim Liu said he saw the edge AI opportunity as “10,000 times bigger than embedded compute”.

Speaking to EE Times Europe once again as part of Embedded World Digital 2021, Liu said the company has continued its focus on edge compute, developing solutions which enable not just edge compute with its AI on modules (AIoM), but also application-ready edge hardware platforms and service-ready edge software.

“Today, edge has become a buzzword,” he said. “Everybody is talking about edge, but there are a lot of ways to interpret what edge is.”

ADLink’s vision is that the move from embedded to edge compute represents a switch from automation to autonomy.

“Individual devices are getting smarter, but it’s not enough,” Liu said. “We need to integrate these smart devices together – make the device, the infrastructure and the operator work together. This is an IoT concept, but today, most IoT applications just think about how to collect the data from the sensor side, and send it from the edge to the cloud.”

This requires a level of connectivity between edge devices. These devices should be enabled to act on incoming data at the right time, which requires real-time infrastructure, distributed among the edge devices themselves.

ADLink Edge compute in the smart factory
Connected intelligent manufacturing devices require real-time infrastructure, distributed among the edge devices themselves

“You need to consider the swarm – swarm autonomy,” he said. “The devices need awareness of each other’s status, and they need to have a small intelligence to make sure they can collaborate.”

This applies to robots, cameras, vehicles and all edge AI applications, which all require reliable, secure real-time networking to share information. Liu strongly believes that a private 5G network is best way to achieve this in smart factories and smart hospitals.

Sitting between the device and private 5G infrastructure should be an edge operating system, which handles data driven networking and resource management. This OS is a vital part of the system which helps integrate the “swarm” of devices so they can work together smoothly.

“The big difference between embedded and edge is how to move from automation to autonomy,” Liu said. “Automation is focused on control, they don’t need a lot of compute resource… moving to the edge requires a different approach, they need to build more and more intelligence inside the device. The second part is connectivity – embedded applications think about the single machine, in a silo. At the edge, this approach is wrong. The whole infrastructure needs to be adapted. Without AI and 5G infrastructure, I don’t think smart manufacturing or smart anything will be a success.”

Smart manufacturing strategy

ADLink’s strategy in the short term is to focus on AI for healthcare and manufacturing, moving to autonomous mobile robots (AMRs) in the medium-term, with autonomous vehicles as a long-term focus.

To this end, this year the world’s number three LCD panel maker, AUO, became a significant shareholder in ADLink. Liu intends to leverage AUO’s factory to build ADLink’s smart manufacturing solution.

“If you want to have a successful business model for 5G and AI, there are three criteria,” Liu said. “You need to have access to the field, you need data and you need domain know-how. Without domain know-how you don’t know how to create good quality data. If you get quality data with your AI training engine, you can create software, but you need validation and verification in the field. So that’s why, no matter whether it’s smart hospitals or smart factories, you need a partner who has the field for your applications.”

On the AMR side, ADLink has partnered with Taiwanese contract electronics manufacturing giant Foxconn. As of 2019, Foxconn’s new chairman Liu Young-Way has given the company three new focus areas: electric vehicles, robotics and medical equipment, according to ADLink’s Jim Liu. The companies have set up a joint venture, FARobot (short for Foxconn ADLink Robot) to develop AMRs for manufacturing, warehousing, retail and medical care.

“The most important thing here is we want to provide a navigation platform that the robot company can base their robots on, whether it’s for healthcare or for different verticals,” ADLink’s Jim Liu said. “We want to provide swarm autonomy to make sure this robot can co-work with the infrastructure. This is totally different to the traditional device-oriented or product-oriented concept.”

As part of the joint venture, ADLink will have access to Foxconn’s factories, data and manufacturing expertise, while Foxconn will benefit from ADLink’s embedded and edge experience. AMR solutions developed by FARobot are likely to be trialed at Foxconn’s factories later this year.

“ADlink isn’t big, but we have some cool technology in hand, and we have 25 years embedded experience,” said Liu. “We know how to build rugged controllers for different verticals for this kind of robotics. We also have special software called DDS [data distribution service] – this was developed for the defence industry, but today it has become a very important technology we call real-time data connectivity.”

Foxconn Chairman Liu Young-Way (left) with ADLink CEO Jim Liu
Foxconn Chairman Liu Young-Way (left) with ADLink CEO Jim Liu and one of the joint venture’s prototype robots (Image: Foxconn)

ADLink’s Cyclone DDS software has been adopted as a key component in the open-source robotics foundations second-generation robot operating system (ROS 2) platform. Cyclone is a data sharing platform for business-critical IoT applications, which is equally applicable to robotics. Robotics can use the software to share data between edge devices in real time, avoiding the cost and delay associated with sending data to the cloud.

ADLink’s long-term strategy includes autonomous driving. The company is a charter member of Foxconn’s open electric vehicle hardware and software platform initiative, MIH, as well as being a member of the Autoware Foundation, which supports open-source autonomous driving projects. This is in keeping with ADLink’s belief in the open-source model.

The company is also a sponsor of the Indy Autonomous Challenge, a competition for universities to develop high speed autonomous vehicles to race at the Indianapolis Motor Speedway. ADLink will provide teams with rugged edge AI hardware, technical resources and engineering support for in-vehicle computing and validation. The race is scheduled for October 2021.

The Indy Autonomous Challenge, sponsored by ADLink
The Indy Autonomous Challenge, sponsored by ADLink, is a competition for university teams to build autonomous racing vehicles (Image: Indy Autonomous Challenge)

Subscribe to Newsletter

Leave a comment