[CRUSH-build] Computer Aided Shooting

[Bill Bennett]

Everybody: try to remember to send technical stuff to build at crush1011.org rather than all at crush1011.org. "all" goes to some people who care about the team but not the technology, so let's not bother them. 

The ideas about driver feedback are great. If the autonomous azimuth adjustments can be made to work, that's great too. The driver feedback will be useful with or w/o the autonomy. You might think about audible feedback so the driver doesn't have to look down at the panel. Or a headset display.

For a side shooter (I'm still favoring this) we have get oriented with the shooter approximately at right angle to the target wall. True for either orientation of shooter of course. With the side shooter, some error in this angle can be compensated by nudging forward or backward to acquire the target from off center, and since we believe we have about a 10 foot range window, this movement generally won't take us out of range. For establishing the 90 degrees, we can use the gyro and/or the camera. Probably the gyro is good to get close, since at very off angles the camera won't see the target, but the camera is best once it starts working, since it's indexed to the real world. The gyro can drift or get wrong due to a collision, etc. 

For getting convinced about the speed at which a tank bot can turn in place, try FWIPER. It's not a perfect comparison, since it only weighs about half what the full size bot will weigh, but it does spin in place very fast. I'm guessing about 1.5 seconds for a 360.
  ----- Original Message ----- 
  From: Gary and Joann Hawes 
  To: all at crush1011.org 
  Sent: Friday, January 13, 2006 5:28 AM
  Subject: [CRUSH-all] Computer Aided Shooting


  This is a completely separate issue from Garrett's campaign to move from a side shooter to a front shooter.  

  As I was looking at the math for using two cameras and triangulation, I  had this idea.  I'm pretty sure we could do it with just one camera.  It requires two lights on the driver's panel.  One light would light to indicate "IN RANGE" and would be based on elevation angle of the camera to the target light and signal strength.  The second panel light would indicate "ON TARGET".  This would be based on the azimuth angle of the camera to the target light and possibly signal strength.  If both panel lights are lit and the driver pulls the trigger, we fire immediately.  But if we are not on target, and the driver pulls the trigger, we allow the computer a finite time to attempt to manuever the robot in azimuth,  either with a horizontal (or azimuth) aimer on the shooter, or by moving the robot itself. ( Back wheels forward and/or front wheels backward turns left, reverse to turn right.)  Assuming the computer is successful, we attain "ON TARGET" and fire.  At that point, the driver can continue to fire.  I think our firing on the fly will only be possible if we have a very skilled driver and/or we have the ability to put an azimuth aimer on the shooter.  What do you think of these ideas?  

  Joann



------------------------------------------------------------------------------


  _______________________________________________
  CRUSH-all mailing list
  CRUSH-all at crush1011.org
  http://crush1011.org/mailman/listinfo/crush-all
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://wizardofaz.whsites.net/pipermail/crush-build/attachments/20060113/3099f1a8/attachment.htm