Co processor Question

Looking into purchasing a co processor for vision detection, any recommendations? +1 if it’s easily compatible with GRIP.

Thanks.

If you want to use GRIP than your best best is an r-pi 2… not sure of the processing power on that though

if you’d like to use CV, than you can go with a beaglebone or TK2

Kangaroo PC

It comes with full 64-bit Windows 10, twice the RAM of a Raspberry Pi, a 32 GB SSD, and an internal battery. It’s a little pricier, but well within the limit for a single component. You can also probably get a 10% education discount.

How would you go about communicating with this? There is no ethernet port on it, only Bluetooth and wifi both of which are illegal to use in competition.

It has a USB port on it… wrong direction though! Connecting it to the RIO would connect two hosts together… which is like plugging a mouse into another mouse. A custom driver for the USB port would have to be written, I guess. Seems like less work to just use a r-pi

Ever looked at this? http://www.em.avnet.com/en-us/design/drc/Pages/Blackfin-Embedded-Vision-Starter-Kit.aspx

The USB on the dock can be used with a USB-to-ethernet adapter (like this one) to plug into the router. Just another option.

Yep, this is most likely what 237 is doing.

We were thinking of doing the vision processing on the Driver Station. The video will be fed to the DS for driving purposes, so no additional wifi traffic.

Is there a co-processor recommendation for use with LabView?

This is really cool. Thank you for sharing.

Odroid c1+, same price as the raspi but also runs linux and has a gpio that is the same as raspi, very good board for the same price.

While all of them will interoperate with a LV programmed roboRIO, at the moment, the only LV tools given to teams are for Windows computers. So an easy onboard processor to target with LV needs to run windows.

The team I’m mentoring intends to do all of our vision on the roboRIO. Coprocessors bring extra MIPs, but also extra complexity and points of failure. So it is worth considering whether the coprocessor is the right approach for your team.

Greg McKaskle

Currently working with the ODroid XU4 with Ubuntu 15.04 32bit.
Mounted in the ODroid default case with 16GB of eMMC.
8GB free for this work and 1.3GB of RAM with Mate running.

Thought the Kangaroo might be a good alternate as well have one on my desk.

The battery in the Kangaroo is attractive. The ODroid’s 8 cores draw 3A at 5VDC (measured with Fluke meter on 100ms max recording mode) with Belkin WiFi, wireless keyboard and a USB camera attached by USB. That is within reach of the old D-Link radio DC/DC converter. Thing is: I worry that a dip below 8VDC might cause a reset so we might put a low drop out DC/DC stepup booster to 15VDC or 18VDC before that D-Link DC/DC converter. That adds volume and weight to the ODroid which would otherwise be smaller than the Kangaroo.

Not sure if we field this frankly. For the last several years I invested in vision related hardware and often it did not get on the robot even if it worked. We are student led so if they choose to not follow through it is their choice. I just provide the investment. Personally I like the modularity of doing the video and camera capture on the coprocessor. If they ditch the idea you just take it off.

The Beaglebone Black is a great coprocessor, with more than enough power for vision tracking.

You might also look at this board:DE0-NANO-SOC.

CPU is a dual core A9 @ 925 MHz, 1 GB DDR3 (32-bit wide) plus FPGA.

DISCLAIMER: I work for Intel (Altera).

Just learned about the Kangaroo. Since it is a Windows desktop, it can be treated much like a DB computer, so RoboRealm, LV examples, and much other SW will run on it and be easily debugged for it.

Greg McKaskle

And you’re recommending an ARM CPU? Clearly they have not made you part of the Intel machine yet!

Heh heh… I got my new badge only just yesterday.

However, I’ve also been told on numerous occasions that our SoC product line (e.g. the Cyclone V SoC on the DE0 Nano SoC board) will have a long life and will have new, ARM-based variants.

Anyhow, I do like this board. I sit beside the marketing guy who oversaw development and was happy to provide him suggestions during that time frame so that it would be useful to robotic applications. This includes the Arduino style headers. There are some things I’d fix on it, but overall, it is a very powerful little guy.

Do you know, off hand, about compatibility with Windows 10 IoT?

How much current at 5V?