Team 81's Autonomous Mode

Here is a video of one of our autonomous modes:
http://robotics.dyndns.org/METALHEADS/Team_81_Auto_Mode.mpg

I have since repaired a couple things like the fact that the turret was aimed wrong for the first ball, etc., but I don’t have any video! :frowning:

Nice work! If you post a video of you ball recognition and gathering that would be AMAZING work to watch.

Unfortunately, our 'bot is on its way to Dreyage right now. :frowning:

I will, however, be putting the ball-finding system on our demonstration 'bot. It can’t pick them up, but it can play “follow the balls”… :smiley:

When it’s done, I’ll post a video of that.

I now have a video of our demo robot chasing a ball (warning: 52 megs!)
http://robotics.dyndns.org/METALHEADS/Team_81_Ball_Finder_Demo.mpg

You will notice the robot consistently placing the ball towards its left-hand side; this is because I had to mount the camera there.

Mounting details here:
http://robotics.dyndns.org/METALHEADS/Team_81_Ball_Finder_Camera_Mounting.jpg

Let me know what you guys think!

EDIT:
This was in full autonomous mode with a PID loop controlling the driving.
There was quite a bit of slop in the gearboxes, hence the “swaying” at the end when trying to line up to charge at the ball.

I havent finished downloading it, but i excpect that this is awsome. Any chance that after competition you will share your code?

All those details still need to be worked out; however I most likely will not be sharing the code.
The FRC interface code possibly, but as I said those details still need to be worked out. :slight_smile:

Offtopic, but we will be releasing all of our code and most likeing our code for our custom control board which we will post pics after it is made. But this is my first year doing FRC coding so nothing to fancy like a ball recognition code. Btw, the file download is slow. Ill try and mirror it for you once i get a copy.

EDIT: Maybe its just me by the download keeps timeing out.

It could be; it is probably the largest file I have hosted yet. I’ll see if I can compress it real quick.

EDIT:
Sorry it took so long. Here is a Windows Media streaming file; it’s only 27mb.
http://robotics.dyndns.org/METALHEADS/Team_81_Ball_Finder_Demo.wmv

AWSOME. Maybe in the off season i will try and get a similiar system up and running.

That is amazing! Did you get it to track the specific colours on the ball? Are you using that code on your robot? how close does it have to be? lol alot of questions, but it was really cool to see!

Interesting. We found that on this year’s camera, when the light was reported to be centered, it was actually to the left (causing it to point to the right). On last year’s, it was reversed (center was right-ward, causing it to point left).

That’s OK, I don’t mind questions :wink:

The system is capable of recognizing and tracking multiple objects of up to 8 different colors under varying lighting and specularity. Distance depends on height; mounted close to the floor like that it could see the ball a good 20 feet away.

I was trying to get it in use for this year’s game, but the autonomous period is so short this year! :frowning:

That problem was CMUCam specific; this is not the CMUCam, so the problem isn’t there. :wink:
However, I did notice that last year, as I had to order a replacement camera. I realized it after the regional, of course! :ahh:

If anyone has any more questions, feel free to ask away! :slight_smile:

Last year, there were many problems with the vision tetras, so how are you getting the camera to recognize non lighted targets? I know even in labivew testing, we couldn’t track someone with a nice bright green bristol board.

This might explain it a bit. This is just using the CMUcam’s camera lens (I believe, don’t quote me on it!), not the board it is normally mounted on, or any of the other stuff, this is directly using it with a Linux “motherboard”.

You are correct. :slight_smile:

I wrote this system from scratch so that I could correct the deficiencies in the CMUCam that I had to struggle with last year. I have found that it does indeed work under a far greater range of conditions than the CMUCam can handle, including recognizing the (quite shiny) balls. I chose to use the OV7620 camera module because I had learned quite a bit about it and its configuration registers through working with the CMUCam.

Hope this clears some things up. :slight_smile:

EDIT:
X-Istence has graciously mirrored the demo video here to lighten the load on my server:
http://personal.x-istence.com/vids/FIRST/Team%2081/Team_81_Ball_Finder_Demo.mpg

Nicely done!

I like the briefcase bot, too :smiley: