Linux-driven dev kit offers three VMs for motor control, vision, and gateway duties
Jun 16, 2020 — by Eric Brown 754 viewsCongatec’s Intel IoT RFP qualified “Realtime Workload Consolidation Starter Kit” for robotics and automation runs 3x VMs including RT Linux and Ubuntu using the RTS Hypervisor on an 8th Gen Conga-TS370 module with an Arria 10 GX FPGA and a Basler camera.
Several years ago, Intel launched an Intel IoT RFP (Request for Proposal) qualification program for IoT kits that use Intel technology for specific use cases like visual retail, smart buildings, security surveillance, and remote health care. The Intel IoT RFP kits include an UP Squared AI Edge Retail Suite from AIM2. Congatec has now joined with Intel to announce the latest qualified Intel IoT RFP, which uses virtualization technology to run multiple OSes on a single system.


Realtime Workload Consolidation Starter Kit (left) and Arria 10 GX equipped XpressGXA10-LP1151AI board from Reflex CES with OpenVINO workflow diagram
(click images to enlarge)
Congatec’s Realtime Workload Consolidation Starter Kit targets vision-based collaborative robotics, automation controls, and autonomous vehicle applications that perform multiple tasks in parallel, including situational awareness capabilities based on deep learning-based AI algorithms. The system consolidates what would potentially be three different computers into one: a real-time motor control computer, a vision processing system, and an IoT gateway.
The consolidated solution is touted for its lower cost, reduced maintenance, and greater security, in addition to the usual edge computing advantages over cloud processing, such as lower latency. “OEMs just need to load their control to the real-time VM and they are ready to enrich their real-time control with data from the situational awareness VM and to communicate in real-time with IIoT/Industry 4.0 counterparts to enable tactile internet controls,” says Congatec. The Workload Consoliation demo was first shown a year ago and then demonstrated again at this year’s Embedded World (see video at bottom).

Realtime Workload Consolidation Starter Kit architecture
(click image to enlarge)
At the core of the system is Congatec’s Conga-TS370 COM Express Type 6 module with an up to 6-core 8th Gen Coffee Lake Xeon E2 processor. The module, which recently added support for 9th Gen Coffee Lake Refresh, also has 32GB of DDR4 and support for Intel Optane and USB 3.1 Gen2.


Conga-TS370 (left) and block diagram
(click images to enlarge)
The CPU hosts the RTS Hypervisor from Real-Time Systems, which here deploys three partitioned virtual machines (VMs), each with its own independent OS. The VMs securely communicate with each other via Ethernet
— ADVERTISEMENT —
One of the six Coffee Lake cores runs the motion control VM, which is equipped with a Servo Motor Pendulum and PLC for situational awareness. It runs Real-Time Linux based deterministic control software. The motor control computer is connected to the Conga-TS370 via USB.
The inverted pendulum is a stand-in for any device that requires precise, deterministic motor control. In the demo, visitors can try to disturb the balance of the pendulum, but the system will reacts instantly to keep the pendulum in balance.
The vision processing stack runs Ubuntu Linux and Intel’s OpenVINO AI toolkit on four of the six cores. This part of the system is equipped with an XpressGXA10-LP1151AI network processing PCIe card from Reflex CES that is built around an Intel Arria 10 GX FPGA with 1150 KLE. The Arria 10 GX is the FPGA-only counterpart to the Linux/Arm enabled Arria 10 SoC.
The FPGA card provides PCIe Gen3 x8, 8GB DDR4, and a QSFP+ cage with 10GbE and 40GbE ports. The XpressGXA10-LP1151AI appears to be a version of the XpressGXA10-LP1151 that is paired with OpenVINO.


Realtime Workload Consolidation Starter Kit demo in May 2019 (left) and at Embedded World 2020
(click images to enlarge)
The unnamed, USB-connected Basler camera is probably a Basler BCON dart. One of the videos says that the FPGA boosts the Basler cameras ability to achieve 30 facial recognitions per second to 300. This is overkill for facial recognition, but useful for machine vision.
There were no details on the IoT gateway that communicates with cloud services except that it offers security features. The IoT VM runs “Edge OS” on the sixth remaining core. In one of the videos, Edge OS is described as “what we call a specific instance of an OS providing gateway functionality.” Most likely, this is yet another version of Linux such as Yocto, but it may possibly be Ubiquiti’s Linux-based EdgeOS stack, which is based on the Debian based Vyatta Network Operating System.
We did not see any detailed specs on the system, which has a glass top for placing objects for AI recognition. It also appears to have a pair of USB ports.
Further information
No pricing or availability information was provided for the Realtime Workload Consolidation Starter Kit. More information may be found in Congatec’s announcement and product page and Intel’s product page.
Please comment here...