Compact, Ubuntu powered robot controllers available in Jetson or Tiger Lake flav...

Compact, Ubuntu powered robot controllers available in Jetson or Tiger Lake flavors

Nov 3, 2021 — by Eric Brown

61 views

Adlink announced two rugged, Ubuntu/ROS driven robotics controllers: a “ROSCube Pico NPN” SBC or box PC based on the Jetson Nano and Xavier NX with 4x GbE and a “ROSCube Pico TGL” box PC based on Tiger Lake-U with 2.5GbE and GbE.

Today at the Robotics and Automation 21 show in Coventry, UK, Adlink unveiled two ROSCube Pico robotics controllers. The ROSCube Pico NPN is available in both board-level and enclosed models, both with a choice of Nvidia Jetson Nano and more powerful Jetson Xavier NX modules. The ROSCube Pico TGL taps an 11th Gen Tiger Lake ULP processor in a box PC form factor and ships with Intel OpenVINO.

ROSCube Pico NPN box PC model (board model shown at top and below) and at right, the ROSCube Pico TGL
(click images to enlarge)

The new ROScube Pico models follow last year’s board-level ROScube Pico Development Kit, available with either an Intel Apollo Lake or Rockchip PX30 processor. Around the same time, Adlink announced an 8th/9th Gen Coffee Lake based ROScube-I robot controller. A month later in July 2020, Adlink announced a ROScube-X system powered by the top-of-the-line Jetson AGX Xavier. Like the ROScube-I, this is much larger than the new ROSCube Pico systems.

At Robotics and Automation 21, Adlink is once again demonstrating its Lidar-equipped, 6th Gen Skylake based NeuronBot robot development and demo kit. Adlink recently announced two autonomous vehicle controllers: the Coffee Lake based AVA-3501 and AVA-3510. Like the ROSCube robotics controllers, they run on Ubuntu Linux with ROS 2 extensions.

ROSCube Pico NPN

Due in large part to the AI algorithms enabled by Nvidia’s powerful GPUs and CUDA libraries, the Jetson modules have been a common choice for robotics development kits. For example, the Jetson Nano has been tapped by SparkFun for its JetBot AI Kit and Waveshare for its own, slightly less feature rich JetBot AI Kit. Like the NeuronBot, the JetBot AI Kits are built around Nvidia’s Jetson Nano Development Kit and include wheels, motor, and other components to build a mobile robot.

By comparison, although the ROSCube Pico NPN is intended for “rapid development and deployment,” it is a robotics controller without mobile robot accessories. It works directly from the Jetson Nano and Xavier NX modules, rather than the official dev kits, and offers more I/O and ruggedization features. The system also adds support for the more powerful, up to 21-TOPS Jetson Xavier NX.

Here is a Jetson Nano vs. Xavier NX comparison:

• Jetson Nano — 69.6 x 45mm; 4x -A57 @ 1.43GHz CPU; 128-core Maxwell GPU; 4GB LPDDR4; 16GB eMMC
• Jetson Xavier NX — 69.6 x 45mm; 6x ARMv8.2 Carmel CPU; 384-core Volta GPU; 48 tensor cores; 8GB LPDDR4x; opt. 8GB to 32GB eMMC

The SBC versions of the ROSCube Pico NPN measure 123.5 x 90mm while the enclosed models measure 140 x 110 x 63.3mm. The Jetson Nano model (NPN-1) has 4GB LPDDR4 loaded on the Jetson module while the Xavier NX model (NPN-2) has 8GB LPDDR4. Both offer Jetson modules with 16GB eMMC 5.1.

The NPN-1 and NPN-2 feature set is the same except that the CANBus header is available only on the Xavier NX-powered NPN-2 board-level product. Major ports include HDMI 2.0, 4x GbE, and 4x USB 3.1 Gen1 (2x with lockable connectors). There is also a micro-USB OTG port for debug and recovery.

ROSCube Pico NPN board-level model
(click image to enlarge)

A microSD slot is available, and on the board-level product it comes pre-installed with a 32GB card. A few other features are available only on the board-level versions. These include a fan control header, 3x LEDs, and a 40-pin GPIO header with SPI, UART, 2x I2C, 7x GPIO, and 10x PWM.

A DB-38 side-panel connector is listed for both versions, but it is possible it is only supplied with the box PC version. The connector provides 2x UART, 2x I2C, 5x PWM, and single SPI, CANbus, extended power on/off, extended SYS reset, and extended force recovery I/Os. On the box PC version, the apparently 38-pin connector is placed in front of the location of the 40-pin header on the board, and it is likely linked to it.

The ROSCube Pico NPN is further equipped with two internal M.2 slots: a B+M-key 2242 for an optional NVMe SSD and an E-key slot for an optional Intel Wireless-AC 9260 with dual-band WiFi and BT 5.0. Other features include a stereo line-out jack, an RTC, and an IMU with 3-axis accelerometer and gyroscope.

A 9-20VDC input jack is accompanied by power and reset buttons and an optional 90W adapter. The system runs at 15W. There is also a power management pin “for extending the function to robots” with PWR_BTN, SYS_RST, force recovery, and power-on LED functions.

The ROSCube Pico NPN has an operating range with 0.6m/s airflow of -20 to 50°C or -20 to 65°C when clocked at 1.2GHz or lower. There is a 95% @40°C (non-condensing) humidity range. The product offers 1Grms vibration resistance compliant with IEC60068-2-64 and 50G shock tolerance per IEC-60068-2-27. The box PC version offers IP40 protection, and both models are listed with a variety of EMS, EMI, and safety compliances.

The box version, which weighs just under a Kilogram, ships with a wall-mount kit. A fan appears to be available with both models.

The ROSCube Pico NPN ships with software including Ubuntu 18.04 L4T and the Adlink Neuron SDK and IDE. Middleware includes ROS/ROS 2, Neuron Library DDS with shared memory, and DDS with extra QoS.

ROScube Pico TGL

The ROScube Pico TGL appears to be the first 11th Gen Tiger Lake powered system specifically designed for robot control. Other rugged, compact Tiger Lake-U box PCs that list robotics among their applications include Vecow’s SPC-7000/7100.

ROSCube Pico TGL
(click image to enlarge)

This box-only product has almost the same design, dimensions, and weight as the ROSCube Pico NPN, coming in at 140 x 110 x 63mm and 1.02 Kg. The system is said to be designed for industrial use service robots such as autonomous mobile robots (AMRs) and autonomous mobile industrial robots (AMIRs) and is touted (without explanation) for its “integrated hard and soft real-time mechanism.”

The ROScube Pico TGL ships with Ubuntu 20.04 plus the Adlink Neuron SDK and ROS/ROS 2 middleware. It also supplies Intel OpenVINO software to tap Tiger Lake’s enhanced CPU/GPU AI acceleration.

As shown in the chart below, the system offers a choice of two quad-core and one dual-core member of Intel’s Tiger Lake-U “E” processors. DDR4-3200 RAM capacity is listed as 8GB (Core-i3), 16GB (i5), and 32GB (i7).

ROSCube Pico TGL CPU and RAM specs
(click image to enlarge)

The system has an M.2 M-Key 2242 slot with PCIe Gen4 x4 for an optional NVMe SSD. The default SKUs offer NVMe PCIe Gen3 SSDs as follows: 64GB (i3), 128GB (i5), and 256GB (i7). You can also choose a 512GB option.

The ROScube Pico TGL is equipped with DP, HDMI, GbE, and 2.5GbE ports. You also get 4x USB 3.2 Gen2 ports: 2x Type A ports, one of which has a lockable connector, and 2x Type-C.

On the side of the system, you will find a COM1 RS-232 port and a COM2 power management port for extending power/reset button and LED functionality to the robot. Other features include dual audio jacks, TPM 2.0, a wall-mount kit, and an M.2 E-Key 2230 slot for an optional Intel Wireless-AC 9260 module.

The ROScube Pico TGL has a 12-19VDC jack with reverse polarity protection and an optional 90W AC/DC adapter. Power consumption runs up to a maximum of 75W.

There is a 0 to 60°C operating range with 95% @40°C (non-condensing) humidity tolerance, Vibration resistance is listed at 3 Grms per IEC 60068-2-64 and shock at 50G, both with an M.2 SSD in place. There are also EMI, EMS, and safety compliances.

Further information

No pricing or availability information was provided for the ROSCube Pico NPN or ROScube Pico TGL. More information may be found in the Adlink announcement on Process and Control Today and the ROSCube Pico NPN and ROScube Pico TGL product pages.