All News | Boards | Chips | Devices | Software | Archive | About | Contact | Subscribe
Follow LinuxGizmos:
Twitter Facebook Pinterest RSS feed
*   get email updates   *

Robotics controller runs Ubuntu on AGX Xavier

Jul 24, 2020 — by Eric Brown 1,302 views

Adlink’s rugged “ROScube-X” robotics controller runs Ubuntu, ROS 2, Adlink’s Neuron SDK, and AWS RoboMaker on a Jetson AGX Xavier with 2x GbE, 5x USB 3.1, 2x M.2, and optional PCIe, CAN, LTE, and time-sync’d GMSL2 camera I/O.

In May, Adlink announced support for Nvidia’s “Nvidia EGX” AI edge solution on four new edge servers using Nvidia’s Jetson Nano, TX2, AGX Xavier, and Tesla. The Xavier-based model, called the M300-Xavier-ROS2, has now apparently been replaced with a similar ROScube-X controller. The system has fewer USB and serial ports, but additional expansion, including support for optional FAKRA connectors for time-synchronized GMSL2 cameras.

ROScube-X (left) and M300-Xavier-ROS2
(click images to enlarge)

Unlike the either discontinued or delayed M300-Xavier-ROS2, the ROScube-X is not promoted as supporting Nvidia EGX, which combines the Nvidia Edge Stack with Red Hat’s Kubernetes-based OpenShift platform. The robotics controller does, however, similarly provide Nvidia’s Ubuntu-powered Jetpack SDK, which helps unlock the Jetson GPU CUDA libraries for AI tasks.

In addition to offering Jetpack, the ROScube-X bundles Adlink’s Ubuntu 18.04 LTS based, ROS 2-enabled Neuron SDK. The Neuron SDK is available on Adlink’s recent, Intel Coffee Lake based ROScube-I robotics controller and Starter Kit. The SDK also runs on Adlink’s 6th Gen Skylake based NeuronBot robot and its ROScube Pico Development Kit, which is available in Apollo Lake or Rockchip PX30 flavors.


Adlink describes the Neuron SDK as “proprietary,” but says it is based on the open source ROS 2 “Foxy Fitzroy” robotics middleware and ships with open source libraries for robot control, including vision, navigation, and motion. The Neuron SDK integrates a hypervisor for safety critical applications, as well as hardware abstraction and drivers for applications such as autonomous mobile robots (AMRs), warehouse AGVs, and autonomous cargo delivery.

The Neuron SDK integrates a tested and optimized version of Eclipse Cyclone DDS. This Adlink contributed Eclypse IOT data sharing standard, which Adlink earlier this month announced with ROS 2 Foxy Fitzroy support, implements the OMG Data Distribution Service interoperability spec.

If that’s not enough robotics software for you, the ROScube-X is qualified to work with Amazon Web Services Internet of Things AWS IOT and AWS RoboMaker. Billed as a “complete cloud solution for robotic developers to simulate, test and securely deploy robotic applications at scale,” AWS RoboMaker is also found on Linux robotics platforms such as SparkFun’s Jetson Nano based JetBot AI Kit and Qualcomm and Thundercomm’s DragonBoard 845 based Robotics RB3 Platform.

Stated Roger Barga, GM, Robotics and Automation Services at AWS: “With the AI compute power of ROScube-X and seamless cloud integration with AWS RoboMaker, developers get a robust toolkit that can help them tackle complex robotic lifecycle management challenges faster than before.”

On to the hardware…

The ROScube-X is built around Nvidia’s high-end Jetson AGX Xavier module, which has powered other robotics solutions including Aetina’s AX710 carrier. The module features 8x 2.26GHz ARMv8.2 cores and a high-end, 512-core, 1.37GHz Nvidia Volta GPU with 64 tensor cores. The module is also equipped with coprocessors including a 7-way VLIW vision chip.

The AGX Xavier variant used here provides 32GB LPDDR4 RAM and 32GB eMMC 5.1. This is accompanied on the ROScube-X by a microSD slot and an M.2 B+M 2280 socket for an NVMe SSD.

ROScube-X with (left) and without the PCIe expansion box
(click images to enlarge)

The ROScube-X has the same 210 x 190 x 80mm case, 9-36V power, and ruggedization specs and many of the same features found as the possibly discontinued M300-Xavier-ROS2. The new specs mention an optional expansion box, but offers no specs Judging from the photo, this appears to be the same expansion cassette found on the M300-Xavier-ROS2, which adds PCIe x8 and PCIe x4 slots.

The key new addition is a pair of optional mini FAKRA connectors, each with 4x camera interfaces, that are said to support up to 8x time-synchronized GMSL2 cameras. The sync occurs with the help of a new 9-axis IMU and perhaps the optional RTC. The SERDES-derived GMSL technology supports bi-directional data, power, and control through a single cable at up to 15 meters without losing latency.

The ROScube-X is equipped with 2x GbE ports and single HDMI 2.0a and USB 3.2 Gen2 ports. There are 4x USB 3.1 Gen1 ports, one of which offers the same lockable connector found on the Gen2 port. Other ports include a micro-USB OTG and both RS-232/485 and RS-232 COM ports.

A DB-50 connector supplies 20x GPIO plus UART, SPI, CAN, I2C, PWM, and isolated ADC. The optional CAN module mentioned in the announcement is not detailed in the specs. An audio I/O jack and 6x LEDs are also available.

In addition to the M.2 B+M storage interface, there is an M.2 A+E-key 1630/2230 slot for optional WiFi/BT (Intel Wireless-AC 9260). A mini-PCIe slot can be fitted with an optional LTE module with the help of the uSIM slot.

The fanless ROScube-X is touted for power consumption that runs as low as 20W. The 9-36V terminal block input is available with reverse polarity protection and power, recovery, and reset buttons. A 160W or 220W AC adapter is optional, and the system ships with a wallmount kit.

The 0 to 50°C tolerant system offers 95% @40°C (non-condensing) humidity protection, 5Grms vibration resistance per IEC 60068-2-64, and 100G shock resistance per MIL-STD-202G Method 213B. There are also certifications for EMI and EMS.

Further information

No pricing or availability information was provided for the ROScube-X. More information may be found in Adlink’s announcement on PRweb and the product page.

Circuit Cellar related posts:

(advertise here)

Print Friendly, PDF & Email

Please comment here...