Onboard Computer

[ekapalka]

Hey! I have a lot of questions relating to this topic. I asked on the rules forum about the legality of having what is essentially a desktop computer on-board a robot, and the answer was something like "as long as each and every part is under $400...". To me, the combination of this knowledge and the upcoming release of the Kinect 2 together are amazing. But something has really been bugging me about this. How can you have an operating system that simply starts up and literally does nothing but execute a single program at startup, with no HIDs or anything? I've seen a few teams using laptops, NUCs, and even bare-bones computers running XP using OpenCV. Do those teams have to start up their robot and processor prior to the match using prehiperals or does it have to boot up with the robot while on the field? I think it might be possible to have a fast boot up and automatic execution using some flavour of Linux...

[androb4]

First do some research on here.

[ekapalka]

I have been. I looked for a while before I decided to ask. If there's anywhere in particular you think I should look, could you suggest it? I know how to do it on a Raspberry Pi, which is about the only on-board computer I have found with any decent amount of FRC documentation, but that's not really what I'm looking for. Thanks, and I apologize if I wasn't very specific.

[edit] This is probably the most informative thread I've found, but it only briefly addresses the issue, with something like "We have a user that auto- logs in on startup and then runs a program in the .preferences". I guess I'll do a bit more research. Thanks

[lynca]

This is a decent thread that discusses using additional processors on an FRC robot.

http://www.chiefdelphi.com/forums/sh...d.php?t=107767

[Jon Stratis]

The way I see it, there are really three options for on-board computing:
1. Use something (like the Raspberry Pi) which is designed for a quick boot, running a single program on start, and collecting/processing sensor data.
2. Use a laptop, with its own integral battery (permitted per R34 in the 2013 rules). This allows you to pre-boot the device, launch any programs you need to, and get things ready to go before taking it on the field.
3. Use desktop components to build your own processing device. You would need to have a quick boot cycle (there are variants of Linux available designed for this), and set it up to automatically run your processing program on start up (Most OS's are capable of running one or more programs at startup/login automatically). As this does not have an integral battery, it would boot when your robot was turned on, very close to the match start time.

I would suggest that #3 is overkill for what we need to do - I can't imagine an FRC robot that really needs more processing power than you can get in a $400 laptop. And while I understand the argument for it, I personally think even a laptop is overkill - a small board (like the Raspberry Pi) should be sufficient for most of what we need.

[ekapalka]

Thank you all so much!

I would suggest that #3 is overkill for what we need to do - I can't imagine an FRC robot that really needs more processing power than you can get in a $400 laptop.

Maybe we'll use four Kinects and have our robot play next year's game fully autonomously :P It'll be for all-purpose vision tracking, not exclusively for being on the robot, so it'll be nice to have around for the programmers. We already have a really slim case that's about half the size of laptop and a 6.7"x6.7" desktop motherboard.

[Al Skierkiewicz]

Don't forget that whatever device you use, it must be firmly mounted to the robot and if it uses some unusual power supply that you add, that must conform to all electrical rules as well.

[protoserge]

This should be a good one.

[billbo911]

There are several low cost single board "PCs" that are capable of performing adequately for offboard vision processing in FRC.

Rasberry Pi has been successfully used, BeagleBones as well.
There are a couple more that come to mind that should be considered.

There is a new Beaglebone Black listed at $45, that looks very promising.
The PCDuino ($60) has been tested by 2073 and is capable of providing tracking data at 15-20 fps. We will be using this and performing further evaluations in off season competitions this year.

Our experience is that the combination of (PCDuino + USB webcam + uSD card) can come in at a cost of about $110. I'm certain the price and performance can be matched or bettered by some of the newer offerings already available and those that will be available soon.

[ekapalka]

Is it just me, or did the words "desktop computer" in the original post somehow become hyper linked to advertisements? If it's not, I swear I didn't do that
[edit] ...what...? It's only happening with chrome...

[protoserge]

If it is Vigilink, I think there was something enabled in the forum not to long ago to get some ad revenue for CD. I thought the mod shut it off since it was so intrusive.

[ekapalka]

Thanks! I was getting worried. In the time since my last post I uninstalled Chrome and started up an antivirus scan... I guess I can stop that now

[faust1706]

Whenever the robot got power, so did our on board processor. We, when we turned on the robot, the vision processor booted up, and the vision program was set to autoboot upon start up, so we didn't have to do anything extra while putting the robot on the field. And the Kinect2 might not be hacked in time for competition next year. Still hoping.