Log in

View Full Version : [video] 2702's successful trailer-hunting autonomous test

Bongle
30-01-2009, 11:49
This is us testing last night with our kitbot. It would have been even better, except the gym has asymmetric lighting, so as soon as the robot got on the left (brighter) side of the target, it couldn't find the target again. We were going to do another attempt where the target zigzagged down the gym, but we got kicked out due to volleyball.

As the YouTube description states, we had someone in control of forward-back motion for safety, but the robot was fully in control of turning left/right. Also, we told the driver to just give it always-on full throttle unless the robot or the guy holding the target might collide.

http://www.youtube.com/watch?v=TiWyO3mFPUI
carbuff
30-01-2009, 11:56
looks like a very nice design
AHS1599
30-01-2009, 12:19
"Stay on target....stay on target..." :D Looking good!
Sean Raia
30-01-2009, 13:09
Nice. we had the same sort of tracking software until our strategy team decided that we shouldn't use it :mad: .
Ian Curtis
30-01-2009, 13:29
That's quite something! That'd be a scary thing to see coming from across the field.

Coach (Blue): "INCOMING!"
Driver (Blue): "I can't get away! It's almost like it's tracking me!"

2702 (Red): *Maniacal Laughter*
Jim Wagner
05-02-2009, 12:29
Was this coded in LabView, if so, did you use the camera tracking demo example? How di dyou incorporate the drive into that example?
Bongle
05-02-2009, 12:43
Was this coded in LabView, if so, did you use the camera tracking demo example? How di dyou incorporate the drive into that example?

We used the base SimpleTracker code (not the more-complex 2-color stuff they posted later) with our own custom "FindTrailerHeading" function that finds green on pink (or pink on green) and returns the distance and camera-relative heading to the target. Then we add that to the current servo heading to get a robot-relative heading, then we feed that to a PID loop which makes the robot point at the targeted heading.

The servo always gets pointed towards the target when we find it, and if it loses the target for 1 second, it goes into a search routine.
MOE
06-02-2009, 01:00
HELLOOooo from the North East :D
Looks like you have been very busy this year, great job well done :yikes: . Hope you have a great year with FIRST this year Good Luck and hope to see Ya.
MOE and Team 88 TJ2 oh yeah ???? OH YEAH !!!! TYE DYE for ever ;)
nathanww
06-02-2009, 01:17
Wow, that's a nice frame rate on that camera.

One thing that you might want to think about implementing is "predictive" tracking--essentially, when the target goes out of frame, instead of just giving up the system calculates the last known vector of the object and then tries to align itself based on that vector and get the object back in frame. This solved a lot of our target-locking problems.

Also, how are you planning on implementing obstacle avoidance/path finding?