double keeper-- yes

[FoleyEngineer]

I hope to see a writeup of the method used to do this.

Jerry w


Hi Jerry,

Hey, I like that "brother's keeper" moniker! Pretty accurate!

To answer your question, here's how it evolved...

Single Score:
First, we got the regular autonomous working. Our original system used the camera and searched until it was locked on target. We took the pan angle and used that to determine how "far off" center our robot was pointing. We then called a "rotate" routine repeatedly until we were within an acceptable error range (pointing pretty much right at the target). We then drove forward monitoring that pan angle. If it got too far off, we'd stop, re-rotate and then continue. All this time, the tilt angle kept increasing as we got closer and closer to the target. When we got to the "right angle", we stopped (or backed up if necessary) and then ran the arm through a pre-cap, pre-spear, spear, down, open, unspear set of position calls. That used to take us about 12 or 13 seconds and then later we managed to speed it up a bit (by moving the arm while we were driving) and then had time to drive back "home" with the arm in the "load wall" position ready to receive a tube at the beginning of the match.

One day this summer we got to thinking that we had so much time left after scoring, that maybe we could try for two. The question was how to do it "legally" and as you suggested we went for the "brother's keeper" technique to use our partner's tube for the second run. Here's how that developed...

Double Score:
First of all, we had to get rid of the rotate, drive, rotate, drive technique for two reasons: One it was too slow, and two, it was too hard to "reverse" to get back "home" for that second tube. We switched to using a "power curve" technique that basically added power to the "outside motors" (based on the pan angle), and had the system drive pretty much full-bore towards the light, monitoring the pan and tilt angles the whole time. Once it reached a "close zone", it literally slams on the brakes for about 1/2 second, and then goes into a "fast creep" mode to get to the right distance from the rack. We then use the rotatable shoulder and elbow to make the final correction for angle and distance and reach out (and over) to drop on the first tube on the middle peg. Then it's race back, again at full speed (in low gear anyway - 7 fps or so) back "home" while simultaneously opening the gripper, bringing the shoulder and elbow back to our "load rear" position. When we get home, we "overshoot" by about 6 inches on purpose to "squeeze" the tube into our gripper and then close the jaws and begin to race back. By the way, we use optical wheel encoders to know where we are and track our progress. Okay, so we race back again at full speed and simultaneously swing the arm and tube over the top running for a different set of arm coordinate points going for the top spider. We usually get there right as the arm does, and we brake again, swing over and down and open the jaws. The whole procedure takes almost exactly 12 seconds every time. That's 4 seconds for drive & score, 4 seconds for return and grab, and 4 seconds for drive and score again. The big variable is how long it takes the camera to "lock on" to the target at the beginning of auton mode. That usually ends up being 3 seconds and is one of the big variables that makes or breaks us, the other being if we "catch" the tube when we come back blind.

In practice, such as the practice area at IRI, we often did it 4 or 5 times in a row in under 15 seconds. That was on our homemade rack however which just wasn't designed to be as exact as the real thing. Throw in the variables of the lights on the field (intensity, interference, color), different rack heights, tube inflation levels, rack rotation and distance, and suddenly you're crossing everything you've got hoping it all works out.

Before this past weekend (the night before in fact), we threw in one little fix that really helped. We added some motor bias correction to steer the robot in reverse back much closer to it's starting point, especially when the rack was at odd angles. This helped us grab that second tube a much higher percentage of the time therefore increasing our chances.

Guess that's about it. Kindof a crazy idea, but one of those things where we said, "What the heck! Why not try it!" and it was fun working it out. Plus the autonomous bonus really made it worth working on. Nothing like the temptation of starting out with a 34 to 0 lead to get your juices flowing!

Hope this helps a bit. Cheers!
John

[GaryVoshol]

John didn't describe how they made a legal pre-match position. They loaded one keeper in their ringer. The other was placed on the floor next to but not touching the bumper. To make this keeper legal, they had one of their alliance members touch the keeper with the corner of the robot. Both keepers were "in possession" of one and only one robot.

As he mentions, the variables were whether the camera could lock onto the light and whether they could go back to find the second keeper. It seemed that the camera locked on quicker if the light happened to be positioned close to a 0-degree angle. If the rack was rotated so that the light was 30+ degrees off, they had a harder time. The first keeper might score, but the delay in finding the light meant they didn't have time to score the 2nd.

Just before that final match, Stevie noticed that we hadn't moved the rack. We went out and pushed, and the light happened to be pretty close to straight-on.

Congrats on finally getting it to go.

[Qbranch]

On the ultra sonics... yes the vex ones are ok... but they use two pins as opposed to one, and because of the dual-pin design they aren't as accurate as single pin units. (I used the Parallax ones)

The good thing was i went off and worked on the ultrasonic on my own... and turned it into a research project. I got it working just before the Championship event... but the team didnt want to try installing it at such a late date.

It came out really amazing... and was highly complex. It used an ultrasonic sensor mounted on an airplane servo, and took 180 degree 'scans' of the area. The robot approached the rack at a set distance (known to be safe, at the limit of legal rack motion) then take a scan. Through a very complex set of digital filters, trend tracking, and logic, the scan data was broken down then all signatures that were identified as pipes (the uprights on the rack) had their angles recorded in an array. Through a fancy little looping structure the pipes (using thier distances and angles) were used to find the closest facet of the rack to the robot.

From there, through a horrendous geometric construction (30 lines of pure geometric calculation alone) a correction angle and from that correction angle, the distance to the target were calculated. Then, the robot held one wheel (PID) and turned the other to the desired angle. Then both wheels accelereated smoothly, then decellerated at the desired correction distance (the PID for the wheel servos were not implemented in season but were completed afterwards, they feature full acceleration control, run speed control, and distance control) and, repeatably, placed a ringer on the rack +/- a few inches. Every time was a close shave... usually took about 14 seconds to place a ringer!

Expect more ultrasonic and servo magic in future 1024 machines...

-q

[Uberbots]

That was amazing. I hate you.... in a graciously professional sort of way.
you guys deserve a snickers bar for that. absolutely amazing.

I wish our robot were legally capable of doing something like this. but... we have a frontload system which makes it impossible to do it quickly

[smurfgirl]

Originally Posted by Uberbots

That was amazing. I hate you.... in a graciously professional sort of way.
you guys deserve a snickers bar for that. absolutely amazing.

I wish our robot were legally capable of doing something like this. but... we have a frontload system which makes it impossible to do it quickly

Maybe if you had written a better autonomous mode... (kidding! Don't explode at me!)

Besides, we have a slightly less cool, but still pretty neat trick. We have that one auto mode setting where we drive around to the other side of the rack and score a keeper over there.

[FoleyEngineer]

Hiya Smurfgirl,

We have a bank of switches that let us pick up to 27 auton patterns. The drive to the other side and score there (high,med,low) was one of them, but we didn't run it in competition for a few reasons. One, by the time we got there, the opponent could easily have already put one on there. Two, we could run into an opponent scoring on a "side" spider and never get there. Three, we could run head-on into an opponent running a defensive pattern (at full speed) and do some serious damage. Overall, we figured it wasn't worth it, but it sure is a cool thing to do! Did you ever get it to go? How often did it get there and score in time? Did you ever run into the opponents?

[Nuttyman54]

I know 1124 ran their backside keeper successfully in a BattleCry 8 practice match. That was amazing to watch!

[Uberbots]

Originally Posted by FoleyEngineer

Hiya Smurfgirl,

We have a bank of switches that let us pick up to 27 auton patterns. The drive to the other side and score there (high,med,low) was one of them, but we didn't run it in competition for a few reasons. One, by the time we got there, the opponent could easily have already put one on there. Two, we could run into an opponent scoring on a "side" spider and never get there. Three, we could run head-on into an opponent running a defensive pattern (at full speed) and do some serious damage. Overall, we figured it wasn't worth it, but it sure is a cool thing to do! Did you ever get it to go? How often did it get there and score in time? Did you ever run into the opponents?

Yeah we did get it to go- about 12 times in practice and once in pseudo competition. As Nutty Said, we did get it to run during a practice at battlecry.

We did it by charging backwards at full speed for 5 seconds, using a drive straight routine that was corrected by the Gyro. (we usually used that for climbing ramps). The routine then ran normally to cap with the desired 8 second cap.

We wanted to use it at battle cry to take advantage of the keeper rule; 3 keepers on the opponent's side would not only break their row but cause a more serious vision impediment to the other team. We ended up not using it, however, because the risk of damage was high, we would start on a side with potentially a lot of defense, the front side autonomous was much more consistent, and the backside autonomous had a failure rate of 7 in 10 trials during practice for unknown reasons.

I would have to say, though, that our coolest trick was the ability to throw a ringer onto the rack. we usually could get a good 2 to 3 yards on that ringer.

[StephLee]

Originally Posted by FoleyEngineer

Hiya Smurfgirl,

We have a bank of switches that let us pick up to 27 auton patterns. The drive to the other side and score there (high,med,low) was one of them, but we didn't run it in competition for a few reasons. One, by the time we got there, the opponent could easily have already put one on there. Two, we could run into an opponent scoring on a "side" spider and never get there. Three, we could run head-on into an opponent running a defensive pattern (at full speed) and do some serious damage. Overall, we figured it wasn't worth it, but it sure is a cool thing to do! Did you ever get it to go? How often did it get there and score in time? Did you ever run into the opponents?

We ran a similar mode, successfully in a match at PARC X and in a few practice matches at BoB. It was a blast to watch, especially if you kept an eye on our programming team when it happened. We were wary of trying it in many matches for the same reason you were, and in the match at PARC, we narrowly missed an opposing robot headed in the opposite direction. The video's on YouTube somewhere, I believe.

[Nawaid Ladak]

from one of the videos i saw, it looked like it was going for the trifecta, am i correct on this?

Great job 910 and congratulations