PDA

View Full Version : replicating the mech warfare environment



muc
10-16-2009, 09:12 AM
Id like to test other IP cameras and camera servers but would need to try and replicate the noisy conditions. What do you all recommend? Is putting a couple of routers close by on the same channel as mine enough? I guess I would also need to have an xbee or two running for the scoring system plus my wireless control link.

Then all I would need is a metal building, few hundred people and a 100 or so robots:rolleyes: But seriously, any help would be appreciated. Id like to setup some kind of test environment.

How did you determine what channel the xbee modules that kept score were set to?

What was the floor like? Was it concrete with epoxy or stain?

How far was each competitors equipment from the arena?

DresnerRobotics
10-16-2009, 10:02 AM
Extremely difficult to replicate such an environment, especially considering the venue is going to change next year. The new venue will hopefully have less noise issues in general though, due to its much more modern construction.

If you're using a Trendnet IP-110W camera, we found those worked extremely well in even the very noisy environment at Fort Mason, so I don't think you need to worry much there.

I don't recall what channel we were using for Xbees, but all of that is modifiable so I wouldn't give it too much concern.

Floor was concrete with some sort of treatment, oil/epoxy/years of tread. Newer venue is supposed to have a similar, but newer floor.

Competitors sit right next to the arena, no more than 20 feet if your bot was on the opposite side of the arena.

lnxfergy
10-16-2009, 11:42 AM
The XBEEs for the scoring system were running on the default channel that they ship at (0x0C). Issy was on a different channel, but I would have to guess that most other XBEEs in use at the event were also on 0x0C -- further evidence that the XBEEs are really good at getting data through even under a congested network.

For the 2010 scoring system, there will be a few changes, I intend to get us off the default channel since most XBEE traffic will be there (and I would guess that most competitors will be using XBEEs for control), and we are going to drop the baud rate, cause we just don't need it so high.

-Fergs

gdubb2
10-16-2009, 12:09 PM
Hey muc,

Good luck with the testing.

I tried some testing a few days ago, in a effort to verify my control system for my mech Bheka. I had a lot of trouble at RG 2009 with the Airlink AIC 250W, so have switched to the Trendnet. Anyway, here's what I did.

In trying to create as much electrical noise as I could, in the 2.4Gig range as well as just general electrical noise, I had (1) Netgear wifi router, (1) laptop using wifi, (2) desktop computer monitors (CRT), (3) Xbee radios (2 on the robot, 1 on laptop), and (1) 1200 watt microwave oven heating a bowl of water.

All of this stuff was operating and located in an area of about 5ft X 10 ft. I then watched the video from the Trendnet camera, and operated the robot via Xbee from the laptop using wifi. I could detect no degradation of the video, or in my control system.

This was just a really simple test, and I have no idea how much leakage the microwave might have been generating, but it was the best I could do with what I have. I did not try the Airlink camera, as it isn't all that great when there is no noise at all.

Gary

Adrenalynn
10-16-2009, 12:27 PM
Hi There,

Which spectrum analyzer/simulator are you using? I can export the captures from mine to a few different formats.

xdream
10-16-2009, 12:58 PM
I think this is the correct thread for this...

What I would like to know is the color temperature of the main arena lighting. I know there was already some discussion on this in the 'encouranging autonomous bot' thread.

If not the color temperature at least what kind of lights/bulbs will be used for the main lighting.

Does anybody know?

Mark

Adrenalynn
10-16-2009, 01:11 PM
>> color temperature of the main arena lighting

In the venue that we were in before - totally random. The area lighting might as well not have existed. The natural light shined through hundred plus year old "coke bottle glass" windows.

The only way was to manually white balance the camera every few minutes. Just 15 minutes saw complete temperature changes. I have the shotlogs here someplace, I could give you the range. It'll be a couple thousand degrees literally though.

That said - I didn't realize the venue was moving an hour away from where it was - so for next year those won't be helpful.

Plan on a combination of random natural lighting + sodium vapor (LPS/SOX). That's fairly unhelpful though since they probably buy whatever LPS they're "getting a deal on this year", and LPS can be had in just about any reasonable temperature...

The takeaway, I think, is plan on manual white balancing in between every match, and plan on the temp. changing substantially during any match.

xdream
10-16-2009, 01:28 PM
Thanks for the info very helpful...no need to send post the range from last year.

I guess there are 2 ways to go here...auto white balance once in the areana on start up and lock in that color temperature..or just leave auto white balance on all the time which can lead to some unpredictable results.

The first way is only good if the lighting doesn't change too much.

Adrenalynn
10-16-2009, 01:49 PM
Add a white balance card on the bot. Point the camera at the card, do a manual white balance, and carry on?

Actually - that's really not a bad idea.

Again - I can't speak to this event yet since I haven't had clarification on which building, the your next challenge is going to be going low-lux enough on a camera you can carry. Are you doing biped or quad?

The Trendnet's that we favor are only 0.5lux. No reason to even consider their white balance in the lighting we've had - there won't be any color reproduction. In last year's venue, I wasn't able to get reliable color from the 0.01lux, 480 line Sony Super Hyper HAD security cameras at street level. My three chip DVCPro ENG camera was able to get pretty good color generally at the sacrifice of a lot of noise. I was sense-up 12 to 18dB. But just my Fujinon glass outweighed any biped there by itself. 23lbs for the camera without battery, transmitters, mics, etc. ;)

I don't expect it to be any brighter at the new venue, probably dimmer through much of the day, although somewhat better controlled.

If Tybs/Fergs want me to light it, we can probably control the lighting pretty well. And if that's the case, I'll be putting down Ushio Halogen through pure white silk softboxes. If we have to overflood an overhead light, I'll be using tungston.

xdream
10-16-2009, 02:23 PM
I can do good color tracking under house lighting tungston conditions...I know what you mean about the lux. When it gets too dark out even the best cameras loose most of their saturation...from my experiments the blue channel goes first but then again it is the weakest channel from the human eye standpoint. Green stays but is to ubiquitous to to pull out the target from the background..that is why I gravitate toward the red channel..Obvously you can mix the channels but I have found I get much better reliability when I try an track around one of the R/G/B values. It doesn't make totally intuative sense to me..my math would tell me that it shouldn't matter but then again the camera is nothing more then R/G/B pixel values so in that way it makes sense.

If someone else can track ANY color as good as one of the single R/G/B channels under variable lighting I would love to see it!

enough rambling for now...

Mark

Adrenalynn
10-16-2009, 03:05 PM
CCDs are, traditionally, most sensitive to green, least sensitive to red. CMOS is traditionally most sensitive to blue, and greens itself all over the place, again, low sensitivity to red. Cameras want to reproduce good images for humans, generally, so they have a lot of tricks up their sleeves, like IR cut filters that notch visible red to keep people from looking as bad as they actually do. ;)

The closer you can get to a pure color, the better chance you're going to have in leaving yourself a broad range around that color and still get enough color information.

So, yeah, I think I can understand and appreciate your experiments assuming I'm correctly reading what you're getting at.

Another thing to consider is the Bayer Domain Scaler (Bayer Filter) in the camera, and how it stacks the pixels.

Not to cross-threads, but another advantage to the targets I posted would be that you could use a monochrome camera and get rid of the Bayer Filter entirely. I can trivially source 600+ line analog cameras in the monochrome with IR and no Bayer or IR Cut filters.

xdream
10-16-2009, 03:34 PM
In case people didn't know CMUcam2 that I'm using on Rover does not do traditional Bayer interpolation...in fact it completly IGNORS the second green pixel!

Adrenalynn
10-16-2009, 04:28 PM
It also has horrible light level ratings unfortunately. I believe it's an OV6620 chipset, right? IIRC, the datasheet on the OV6620 specs it at 3 Lux. Orders of magnitude above what we're looking at for light levels.

Adrenalynn
10-16-2009, 04:43 PM
I just put a call in to a friend. He actually did some paid Maker Faire photography, so he'll have light meter readings for some of the buildings. See what they actually look like. Also, we should be able to extract color temp info from his Camera RAW white balance shots. [I was shooting the Point-and-Shoot. No way I felt like dragging the big camera around]

muc
10-16-2009, 05:24 PM
Gary, my tests will be pretty much the same way. Just add noise until I can see degradation in the video. I need to get the trendnet camera and use it as a benchmark. Im still waiting for it to go on sale.

Adrenalynn, I dont have any high end equipment. The only thing I use to view the 2.4GHz spectrum is an older version of the wi-spy from www.metageek.net (http://www.metageek.net/). It would be cool to see some pictures but they would need to be jpg or gif.

Mark

xdream
10-16-2009, 05:33 PM
That would be awesome Adrenalynn!

Adrenalynn
10-16-2009, 07:10 PM
3200 to 3700K depending upon the building and the run-time on the bulbs (and the light leakage from outside). That makes me think they're atypical and running metal halides... Less efficient, more expensive [shrug] - I'm an electrical engineer, not a building engineer. :)

I haven't personally confirmed that - I'm waiting for this evening when I'm supposed to receive some Camera RAW images for my own testing.

Light meter was pretty consistent at 0.7 - 1lux. Somewhere in the twighlight neighborhood, with the exception of the Dark Room, of course. That also seems to jive with what I observed with the 0.5lux Trendnet cameras. Dim, noisy, but functional [in that other venue].


Really, we need to keep our eyes peeled for a near-term event there and bring in the real gear to get a better idea.

xdream
10-16-2009, 08:56 PM
Wow that range converts to about -2EV which is the lowest my meter can read. When I present that light level in my lab even I'm having trouble seeing well with my own eyes let alone through a 2 or 3 lux camera.

From your previous comments I'm assuming all cameras were pushing their minimum lux rating and working hard under those conditions. +3EV (20 lux) is still kind of a dim evironment to my eyes.

I hope we get a little more illumination for the 2010 event!

Mark

Adrenalynn
10-16-2009, 11:12 PM
It's twilight, which most people don't have too much trouble seeing in. My cameras were able to reproduce weak color.

A full moon is an order of magnitude less light. Starlight/city light glow is an order of magnitude less than that, and a cloudy night is an order less than that. Below that is 0 lux. As I mentioned - the Trendnets weren't able to reproduce color well or consistently at 0.5 lux, and 3 lux would be blind.

If you scrap the CMU cam and go to a Sony ExWave HAD, or better, the new Sony Interline Transfer [ExWave 2] chipsets you'll be fine for ambient light. (0.001 and 0.0001 lux)

xdream
10-16-2009, 11:21 PM
Thx...worth investigating for sure!