View Full Version : Thinking about DefConBots with roborealm

06-20-2008, 02:42 PM
bah. Just lost my long post...

So anywho, I'm starting to think about defconbots (http://www.defconbots.org) and using roborealm. The vision processing is proving trivial, even though I'm at an oblique angle in these shots and 3x farther away, but I'm not really sure about positioning the gun(s) yet.

Since Alex and Tyb were talking about Roborealm (http://www.roborealm.com) I figured I'd post these shots. If anyone wants the .robo, let me know. I'd welcome any thoughts about using this to postion the guns too.

It's only taking about 16ms total to process this on a EEE PC. A pink one at that. ;)

06-20-2008, 02:58 PM
I stopped reading when you proclaimed you have a pink EEE :p


I have an auto-aiming .robo in the works using a laserpointer for additional accuracy in mind for my mech :D I'm using your .robo you shared as a rough guideline... still need to figure out how I can override it with manual control on the fly though.

06-20-2008, 03:06 PM
I have to be allowed SOME cuteness. It's a rule.

I'm really of two minds. Ok, three minds... :P

Camera on the gun scanning for targets?
Camera overlooking the field and directing the gun?
Combination of the two?
Camera + Laser range finder looking over the field?

Wait - that's four. Amongst the methods I'm considering are such diverse elements as...

06-20-2008, 03:13 PM
Are we talking for Mech Wars?

I would think any sort of tracking system would need to be self contained on your bot, no overhead cameras allowed as I'm trying to stick with a POV from the mech only scheme.

06-20-2008, 03:23 PM
Sorry, no, I was talking about DefConBots.

06-20-2008, 03:30 PM
I have to be allowed SOME cuteness. It's a rule.

I'm really of two minds. Ok, three minds... :P

Camera on the gun scanning for targets?
Camera overlooking the field and directing the gun?
Combination of the two?
Camera + Laser range finder looking over the field?

Wait - that's four. Amongst the methods I'm considering are such diverse elements as...

If your second choice is nothing more than a camera on a pan and tilt platform with a gun off to the side, then you could just consider your first option to be a pan and tilt camera platform with the gun attached. Just cuts back on the number of servos you have to use. Or is the second option use a stationary camera? If you do the second option then you should use a laser.

06-20-2008, 03:31 PM
That looks like an interesting competition....

06-20-2008, 03:41 PM
It would be a fixed camera, sthmck. Which provides invariable X,Y and target dead or live, even if one bounces back up.

Doesn't it though, Sienna? So much software problem solving opportunity that isn't locked into tough and expensive hardware.

06-20-2008, 05:14 PM
Hmm, Defconbots. I gave up on that one due to overabundance of other projects (I am trying to concentrate on Deskpet) and the fact that some guys at work (yes, here in Davis) are building one too awesome for me to compete with. I would give details, but they've sworn me to secrecy ;) One of the team members is the guy who ran the defconbots challenge at RoboGames this year.

Here's a link to the prototype pan & tilt that I was constructing for it. I attached a laser pointer and used it to keep the cats entertained.


Back when I was working on it, I had started to come to the conclusion that the vision system would be the easiest (I was going to use OpenCV (http://sourceforge.net/projects/opencvlibrary/)), followed by the control system, with the actual mechanics being the hardest. It needs fast strong motors, with very little backlash in the geartrain. I'll be interested to follow your progress.

06-20-2008, 10:30 PM
Cute video!

I'm actually going at it from a different direction. I have the mechanics nailed down pretty well. A little less traditional. Just trying to figure out my optimum path planning for my non-traditional method. :)

06-20-2008, 10:41 PM
I like or favour the camera tracking on the weapon system itself, but that's just my opion and I am open to all other solutions.Nice vid rbotguy.

06-20-2008, 10:54 PM
You didn't provide a reason? I can think of a handful of reasons not to do that, so if there's any good ones for it, I'm all ears!

06-20-2008, 11:24 PM
It makes for an easier targeting system to have the camera follow the weapon system, no different from a sight on a ground or anti aircraft system, you could try a cross hair on the camera an tune in your weapons to a pre determined range say 10-20 ft what the camera see's is where the weapon fires. But as i said I'm very open to other suggestions on this.

06-21-2008, 12:14 AM
That's great except for the path planning being far less intelligent, and if you're looking over here and the target vibrates back on its feet over there, you won't see it. It also doesn't allow for multiple weapons working in cooperation.

I appreciate your thoughts and views and input - that's why I posted!

06-21-2008, 03:08 AM
Ah, I see your point Adrenalynn, Thank you for clarifying that. Back to the drawing board. So where do you advise fitting a camera for a weapons system on a bot? Head,shoulder torso.

06-21-2008, 11:24 AM
I'm still not sure it's *not* the "best" method... It's one of those things that has to be worked through, otherwise you're just duplicating what's already been done.

06-22-2008, 11:33 PM
I was thinking about the DefConBots competition and went to Robogames08 to observe the demo rounds there. I took some photos and wrote some basic observations that I thought I'd share here.

There were three designs there, and sadly I didn't catch anybody's name, so I'll just refer to them by the number that I think they eventually placed in. My apologies if I've gotten any of the details wrong.

http://www.flickr.com/photos/[email protected]/sets/72157605765999733/

I do agree with your earlier comments about the benefits of a static camera vs the gun inline with the camera. I think a static camera can be a more stable one, allowing for higher resolution images and the coordination of multiple weapons. But the winning designs seemed to go for the inline arrangement, so perhaps the jitter and not-seeing-the-whole-field problem is not as bad as I thought.

The other item I thought that was important was to minimize the amount of moving mass. A lighter gun means that less powerful servos can move it quicker with no slop.

BTW, the winner hit all 20 targets in like 24 seconds. I think he told me that the winner at defCon last year did all 20 in 20 seconds.

I'm hoping to build some simple demo models in the next week and a half.

06-23-2008, 04:20 AM
I'm with you on several points, except for the "'cause the winner last year didn't do it, we don't need to do it". I think my goal would be to innovate.

I think you're correct in your other observations, though!

I do have video I shot of the 24sec round, if you'd like to review it.

06-23-2008, 05:53 PM
I don't mean to say that "since the winner didn't do it, we don't need to," Just pointing out that the more successful designs seemed to use an inline camera. But I'm sure if the 3rd place finisher (with the plywood) had minimized the moving mass of his gun and pan/tilt harness, his static camera would have been fine.

I am toying with two gun designs. One is a compressed air ping-pong ball launcher that uses a solenoid for the pressure release to fire and a servo to rotate the firing chamber 90 degrees to load the next shot. I have no idea how long it will take to get pressure back up, and I worry that the entire firing mechanism will be to heavy to aim quickly. Obviously I need to do better than one accurate shot per second.

The second design was modeled after a tennis ball or baseball launcher, in that there are two wheels turning in opposite directions, with the projectile run between them. Also heavy, but should allow for a very good rate of fire.

I just can't bear to only use a modified airsoft gun. I'd like something slightly more original.

We'll see what I can get built this weekend.

06-23-2008, 11:21 PM
Hi again, Lamont - welcome to the TRC, if I haven't already welcomed you properly!

I really appreciate your input, thoughts, and insight.

I'm heading off in the multi-gun direction, still airsoft, but custom-built. I have several projects that absolutely *must* get finished before I can even think about starting a build. My game plan right now involves competition barrels 6.04mm paired with probably actuated gas release mechanisms.

Still the question remains - single static, multi-mobile, or combination design. Plenty of time to engineer that out, huh? The difference is the spice and all.

I'd be concerned about the irregularity of pingpong balls. I've never really been able to get 'em to go where I think I'm putting them. That and mass-vs-drag. Thoughts?

06-24-2008, 03:12 PM
Why thank you, I do feel quite welcome. I'm excited to be among like minded nerds.

I'll probably end up using airsoft pellets as well, but I'm still playing with the compressed air pingpong cannon. I did some research into barrel rifling to impart a longitudinal spin to the ping pong ball. I was going to use a CNC to carve the firing chamber and barrel out of seaboard (a cheap plastic that mills really easily), then paint half a ball back and see if I could measure the spin with a high speed camera. (I don't have a camera that can do anything faster than 30fps, so that's probably out unless I can borrow one of those fancy high-speed canon cameras) But maybe with a good barrel and some baffles on the side for the air to escape, I won't even need that.

So far nobody is using opposing spinning wheels to launch projectiles (that I've seen), so I'll also give that a shot to see what kind of accuracy I can get.

Meanwhile, the organizer from defcon put up his RoboGames writeup and some photos. I'm in the top photo, the left-most person facing the action. (black hoodie, not facing the camera, brown hair, glasses)

06-24-2008, 06:23 PM
I have to get on the plain, but I'm the tall woman, short dark hair, brown blazer, cream colored top, jeans, on the photo-right side next to a shorter scruffy guy. :)

07-10-2008, 03:59 PM
I am still waiting for my airsoft tank cannons to arrive. dcalkins and I have a date to experiment on them, but it's frustrating to not have them in my grubby little hands yet. Especially since at this point I think I'll end up scrapping the guns as delivered and just using their design to make something a little tougher. ebay has a whole bunch of upgrade parts (replacement pistons, gears, springs) or I might hit up electricity to see what his store sells.

But before I go about building something new, I'd like to at least be able to test what I've got. Stupid slow boat from china (well, hong kong, but you get the idea)

07-10-2008, 04:31 PM
Stupid slow boat from china (well, hong kong, but you get the idea)

Yah, those guys should really start rowing faster!
or, as my boss would say, WORK HARDER!!!

iee.. sorry bout that

07-10-2008, 04:32 PM
Mine actually got here surprisingly fast, but I did pay for EMS Air... I think mine took five days inclusive of a weekend.