Onboard computer for UAV image processing...

N

-=Nu||=-

Gawd
Joined
Jan 5, 2002
Messages
974
Hi everybody,
I'm involved in a project to build an autonomous UAV (Unmanned Aerial Vehicle) to compete in a competition with the goal of locating a lost hiker in the Australian outback. We are looking for a small, light, and powerful system to do our image processing.

I had previously hoped a Nano ITX system would be powerful enough, but from what I have read the VIA 1GHz Nano ITX systems are somewhat weak. The only thing holding us back from choosing a Mini ITX system is that it might not actually fit on the plane, and it is slightly heavier. I will take measurements of the plane and see how big it can be later this afternoon. Power consumption is also an issue, since it will be running off of a 14.7V 3.6Ah (amp hour) battery (subject to change).

I'm hoping there are some systems slightly smaller than Mini ITX with an Intel of AMD processor, but I wasn't able to find any. Anyways let me know if you have any ideas. Thanks in advance!
 
3.6A x 14.7v = 53w? Ouch. miniITX is 170x170mm or 6.7” x 6.7” so hopefully for your sake it fits. There are SBC solutions out there, but you are really going to have to look for them and when you get that small, the biggest problem is processing power. If a VIA is not powerful enough and there is a good chance it is not (1GHz VIA = approx .25GHz PIII/Pentium M), the only chips that might do what you need with that low power would probably be ULV Pentium Ms or AMD Geodes and both are going to be expensive.

I would check out www.ibase.com.tw. They have two P-M and Geode SBC solutions that may work for you. Send them an e-mail - they will sell directly to you.
 
Thanks for the links, we are looking into other batteries with more capacity and less weight, but they will only need to last for 1.5 hours. It looks like we can have maximum dimensions of about 6"x10"x2.5"...

EDIT: Ibase seems to have some really nice systems, I think what they have may work, hopefully it's not ridiculously expensive...
 
Give you an idea, I bought one of their P-M boards and it was around $300. Now that same board runs $200. ibase is "cheap" in comparison to more name brands. No idea what the SBC boards run, but for sure they will probably be more expensive. Smaller you get, the more niche you get and it’s hard to avoid having those prices skyrocket…

Sounds like a cool project though. ;)
 
Wondernerd said:
How accurate do you need to be? It can be done
Click to expand...

That's pretty sweet, I didn't read the entire thing, but I'm not sure how fast they needed to process the images, we will be needing to process them at 30fps, I'd rather go for overkill than be underpowered.


sleepeeg3 said:
Few other manufacturers of SBC systems. There prices may be more reasonable:
http://forums.extremeoverclocking.com/showpost.php?p=2656524&postcount=1
Click to expand...

Thanks for all those links, I will definitely check those out.

_G_ said:
what about an ultra portable laptop?
Click to expand...

Our maximum dimensions are 6"x10"x2.5", so I'm not sure a laptop would fit, and laptops also have the cost of LCDs and enclosures associated with them so they are a bit more expensive than a single board computer.
 
This was posted on the Casemoding forum by Widefault:

Commel LS3-71

The specs look the same as the iBase (they use the same G945 chipset):
Core2 Duo / Core Duo / Core Solo processor @ 533 / 677 MHz FSB.
Memory: 1 x 200-pin DDR2 SO-DIMM up to 2.0 GB.
CompactFlash Type-II socket for solid-state HDD (better shock resistance)
lots of other normal stuff: LAN, sound, video, etc.
On-board 8-bit I/O (might come in handy)
Form factor: 3.5" Miniboard = 146 x 101 mm (L x W)
and best of all, 8 ~ 24V full range DC input.
not sure about the price, $300? $350?

Visit the webpage for more specs and other "Miniboard" products
 
You really should demo the some code out before you start thinking about hardware. What you want to do might not even be feasible with x86 hardware. It all depends on what type of image processing you're doing. Is it live video you'll be processing? Are you just converting still images? You need to prototype your software before you even start debating what hardware to run. Other things such as storage media & OS should also play a part in your hardware decision.

If you want to test the VIA part look on eBay for a Via C3 or Epia in a larger form factor. They run as low as $50. Even if you run it on a slower part(800mhz for example) if it's pegging the CPU @ %100, the extra 200Mhz a 1Ghz CPU would provide probably isn't going to help.

Power too, you're probably going to need more juice. A Core2(laptop part @ 35W) will eat through that battery in under 2hrs. How long does the UAV need to be running this software? Will it be running CPU intensive software all the time? Could a switch be devised to turn the computer on only when needed?

More info needed before you start nailing down what hardware to get.
 
Is it absolutely necessary to process the data on the UAV itself? Why not transmit the raw data and process at your ground station? Remember that the pdf linked above is a simple vision system, not something trying to pick out a person from ground clutter.

The power-size-weight budgets just don't add up to be able to process on your aircraft, especially once you factor in propulsion and flight systems (GPS, autopilot, control servos), your camera system, and any transmitters/receivers.

I'm guessing that you want your camera to key on a person picked out from the ground, update the autopilot to keep the object in sight, transmit an image/video to a ground station so an operator can say yea or nea and send the drone on its way.

AFAIK even advanced systems like the predator rely on operator commands to key on a target.
 
AFAIK even advanced systems like the predator rely on operator commands to key on a target
Click to expand...

This is absolutely correct. I've been supporting two different UAV projects for the last two years and both require a ground station. The UAV's purpose is to provide an airframe for the camera and signal transmitting equipment. All video processing is done on the ground.
 
Shadowspawn said:
This is absolutely correct. I've been supporting two different UAV projects for the last two years and both require a ground station. The UAV's purpose is to provide an airframe for the camera and signal transmitting equipment. All video processing is done on the ground.
Click to expand...

Somebody asked above how long it will need to run, the UAV will only need to fly 1 hour at a time, so 1 hour's worth of batteries should be feasible.

I have considered doing the processing at the ground station, the only thing that was holding that back, was the fact that the video will need to be sent back over a distance of 7 miles at the most, and I was having trouble finding a wireless link that has that kind of range and bandwidth. I was considering having a 802.11g link with a high gain antenna on the UAV and a parabolic dish on the ground, but I'm not sure what kind of range we can get with that. Any ideas or suggestions for such a long range wireless link? And yes it will be open and flat terrain, so essentially LOS (line of sight).
 
You can't use satellite technology, or does it need to be done via radio waves?
 
protias said:
You can't use satellite technology, or does it need to be done via radio waves?
Click to expand...

Satellite would probably be quite expensive and could be rather slow due to the high latency. So radio waves would be the only feasible way that I could think of.
 
-=Nu||=- said:
Satellite would probably be quite expensive and could be rather slow due to the high latency. So radio waves would be the only feasible way that I could think of.
Click to expand...

Well, the military uses satellite, hence my suggestion. How "real-time" does it need to be?
 
protias said:
Well, the military uses satellite, hence my suggestion. How "real-time" does it need to be?
Click to expand...

I guess whether or not it's "real-time" might not matter that much, as long as it's also sending the GPS coordinates along with the images, which it will be doing. But I'm pretty sure a satellite link would be prohibitively expensive, at least in our situation, I'm sure the military can afford it though :).
 
Consider this:

A 5 minute backup battery weighs somewhere around 15 pounds. You need to run the computer for an hour (even a low power one), plus your other big draw, propulsion.

Most electric R/C aircraft have flight times of about 10-12 minutes, and they don't haul computers around. Remember, as weight increases, power requirements increase, as you will need to produce more lift to get the plane off the ground, which means increasing battery weight to keep the airplane in the air.....you see the circular problem.

Lift Force Required=mass times gravity (roughly)
Lift = 1/2 rho * S * V^2 * CsubL

So to increase lift, you increase the velocity (more power required), the wing area (more drag, more power required, or the coefficient of lift (which increases drag as it increases, requiring more, you guessed it, power.)

Also consider crashes. One thing any R/C pilot will tell you..."You WILL crash." Can your computer system survive a crash? Repeated crashes? Are you willing to risk several hundred dollars in equipment?


As for data transmission, you are probably going to have to look at licensed equipment (HAM radio stuff) to accomplish what you want to do. Most unlicensed equipment just isn't able to put out the required power.

The way I see it, you have several large problems to address. The first is your mission: Lock down exactly what the requirements of your mission are. BE SPECIFIC! The second is your propulsion system: Do some research on what an electrically powered aircraft (of the size you seem to be indicating) is capable of. Don't assume or guess, the data is out there to figure all this out (Motor capabilities and their power requirements, power densities of differing power solutions, etc). I think you will find that keeping an electrically powered aircraft in the air for over an hour is going to be a complicated project in and of itself.

Also, have you sourced the R/C autopilot/GPS modules? I know that some of them, SBC's with solid state accelerometers can run into the several thousands of dollars.
 
[RCKY] Thor;1030789121 said:
Consider this:

A 5 minute backup battery weighs somewhere around 15 pounds. You need to run the computer for an hour (even a low power one), plus your other big draw, propulsion.

Most electric R/C aircraft have flight times of about 10-12 minutes, and they don't haul computers around. Remember, as weight increases, power requirements increase, as you will need to produce more lift to get the plane off the ground, which means increasing battery weight to keep the airplane in the air.....you see the circular problem.

Lift Force Required=mass times gravity (roughly)
Lift = 1/2 rho * S * V^2 * CsubL

So to increase lift, you increase the velocity (more power required), the wing area (more drag, more power required, or the coefficient of lift (which increases drag as it increases, requiring more, you guessed it, power.)

Also consider crashes. One thing any R/C pilot will tell you..."You WILL crash." Can your computer system survive a crash? Repeated crashes? Are you willing to risk several hundred dollars in equipment?


As for data transmission, you are probably going to have to look at licensed equipment (HAM radio stuff) to accomplish what you want to do. Most unlicensed equipment just isn't able to put out the required power.

The way I see it, you have several large problems to address. The first is your mission: Lock down exactly what the requirements of your mission are. BE SPECIFIC! The second is your propulsion system: Do some research on what an electrically powered aircraft (of the size you seem to be indicating) is capable of. Don't assume or guess, the data is out there to figure all this out (Motor capabilities and their power requirements, power densities of differing power solutions, etc). I think you will find that keeping an electrically powered aircraft in the air for over an hour is going to be a complicated project in and of itself.

Also, have you sourced the R/C autopilot/GPS modules? I know that some of them, SBC's with solid state accelerometers can run into the several thousands of dollars.
Click to expand...

Well as far as the aircraft we already have that, it's an 80" wingspan Sig Kadet powered by a 0.70 OS four-stroke engine, and we also have an Aerospace Engineer working on the project and he's done the calculations for the airframe and the weight it should be able to carry, so I'm not too worried about that.

The 5 minute battery backups use lead acid batteries usually, we are looking at NiHM or Lithium Ion, which has a significantly higher power density. Currently have a 14.7V 3.6Ah battery pack which should be enough power for the hour that we need to stay in the air, but if needed we might go with Lithium-Ion or LiPo. We also already have the autopilot selected, and hopefully we will have some nice strong enclosures for the electronics so in the event of a crash they will survive.
 
You must log in or register to reply here.
Back
Top