Team Update #18

[MrForbes]

Originally Posted by Brandon Holley

If thats the case, then why did the first two weeks of the season use manual scoring to determine minibot race winners?

That's a very good question. I guess the officials didn't understand the rules.

Originally Posted by Karthik

If the issue is that teams aren't contacting the top target for long enough, I'm more concerned as nowhere in any official specification or guideline have I seen a duration of impact clause.

If we'd all had our act together two months ago, we would have asked

[colt527]

Originally Posted by Karthik

Ken, what would be considered bouncing off too fast? What is the duration that teams need to contact the disc to ensure that the target is triggered?

I do not have a answer for this. So far the only thing I heard of was the speculation of 75ms floating around in this thread. I can see if I can find an answer to this later.

It makes sense to me that if it turns out there are bots that definitely don't hit the bolt and hit the pad hard enough and still don't trigger that duration would be the next most probable cause. I can imagine a pretty violent bounce on some minibots. I would like to believe FMS would still catch those, just based on some of the minibots I saw at Pittsburgh which bounce pretty good that got picked up, however I can not say that bouncing too quick is an impossible cause of a false negative.

I think teams can alleviate the duration problem pretty easily as well by adding some cushioning or something, but I know a real hard number for duration time would help team immensely.

[JVN]

Originally Posted by squirrel

I put more credence with those who read the rules and understand what the rules require the TEAMS to do.

There is no "should have TRIGGERED the TARGET" in the rules. You do it, you get credit. You don't do it, you don't get credit for it.

Part of the design challenge is to make sure your MINIBOT TRIGGERS the TOWER. Some TEAMS seem to have missed that. Apparently they thought the challenge was to get a minibot up the TOWER as quickly as possible.

The challenge was to get the minibot to trigger the target as quickly as possible. This is an optimization problem. Many of us used the specifications given as we worked out the optimization.

As they say in Animal House: "You effed up! You trusted us!"

To imply that all teams should have built a slow super-overpowered minibot just to ensure they trigger the target if the FIRST specs were wrong is silly. We (as good engineers) should have worked within the specs given to us.

BTW - I have no idea if there is a problem or not, but I'm surprised by your dismissal of the fears of many by saying "you should have built your minibot to hit harder, not to the spec."

-John

[MrForbes]

Apparently not many (none?) of us realized the spec was incomplete.

I sure missed it.

[Kims Robot]

Originally Posted by martin417

In FIRST, the manuals and rules are the specs... The peg heights are not in the rules either. The entirety of the game documentation is the specification document to which all teams work.

I beg to differ... the peg heights are in GE1100-1106 and are referenced on page 3/10 of the 2011 FRC Game Manual, Section 2 - The Arena. Yes they could have been better spelled out. In the "real" engineering world, engineers reference other documents all the time, rather than copy/pasting specs into every document they make because it reduces the risk of forgetting to update documents when a change is made. GE-11036 tells you exactly how to build a sensor plate, yet how many teams actually did it?

Originally Posted by colt527

1.) Hit the bolt
2.) Did not hit with enough force (like, creeped up really really slowly with the wheels slipping)
3.) Turned off / reversed immediately after hitting the bottom plate without pushing much at all

Bingo!

Originally Posted by BrendanB

IF the triggering programs and times of triggering on the towers hadn't been changed by FIRST between week 1 and 2 then this wouldn't be a problem for me. What is a problem is that it is changed to a level of "it seems to be working" instead of "every minibot who goes up the pole and exerts the 2-4N of force will trigger the top".

And to top it off humans wont' be used as the backup when electronics fail... seems backwards to me.

I am not sure what you call "it seems to be working" as week 1 they CLEARLY weren't working. There were a ton of false triggers (robots hitting the tower), and then when they implemented a duration, some robots were too fast, so they resorted to manual scoring for week 1. So FIRST fixed them. They did not change the plates, just the electronics/sensors.

In my mind there is absolutely no way a ref or any spectator can tell for certain that a minibot has put enough pressure on the plate, not hit a bolt and reliably triggered the top. The field crew will be able to test & determine if the tower sensors are working, and if a sensor breaks, I am certain they will fix it.

Originally Posted by Karthik

Ken's point is a very valid one. Does anyone have any video of Minibots hitting the top target but not triggering the target during week 2? Right now we're just dealing in speculation which isn't leading to anything productive. From what I saw in Pittsburgh, the large majority of minibots were in fact triggering the top target. If this wasn't the case at a regional you were attending, let's try and figure out what the issue was that was causing these false negatives.

If the issue is that teams aren't contacting the top target for long enough, I'm more concerned as nowhere in any official specification or guideline have I seen a duration of impact clause.

EXACTLY... I have yet to see a single video that convinces me otherwise. Everyone can do all the math they want, but without factoring in things like when does your robot reverse direction, what is its deceleration,etc, its impossible for me to conclude that the towers aren't working.

And I agree with everyone saying there is too much speculation in here... has anyone actually gone into the Q&A and asked the question? Asked what the sensor is? what the duration is? Your team leader has access to the Q&A forums... USE THEM.

[martin417]

Originally Posted by Kims Robot

I beg to differ... the peg heights are in GE1100-1106 and are referenced on page 3/10 of the 2011 FRC Game Manual, Section 2 - The Arena. Yes they could have been better spelled out. In the "real" engineering world, engineers reference other documents all the time, rather than copy/pasting specs into every document they make because it reduces the risk of forgetting to update documents when a change is made. GE-11036 tells you exactly how to build a sensor plate, yet how many teams actually did it?

You missed my point entirely. Squirrel was making the point that the 4N number was not in the rules (one of those things with the <Gxxx>), but in the arena drawings. My point was that the arena drawings ARE part of the specs and are just as important as the rules (<Gxxx> again).

Also, even if a team built the sensor plate, it is apparent that that is not enough. You have to know how that switch trigger will be detected. What algorithm is used to reject pole strikes by robots? how long does the switch have to be depressed? etc.

[Kims Robot]

Originally Posted by martin417

Also, even if a team built the sensor plate, it is apparent that that is not enough. You have to know how that switch trigger will be detected. What algorithm is used to reject pole strikes by robots? how long does the switch have to be depressed? etc.

Ask in the Q&A.

Sorry if I misread your original post... I'm a little annoyed that everyone is blaming FIRST for this, when they are clearly trying to make an improvement over week 1, and no one can move that a minibot that should have triggered the pole didn't.

[IKE]

There was a question in the Q&A asking if the bot had to stay up for any period of time.

The response was that it just had to trigger the plate.

As far as padding goes, we tried this in week 1 with 1/2" of "grip-pad" and electrical tape formed into a cushion. This did not improve the reliability of triggering in week 1 at all.

I measured the force our light switch requires to activate. It was 5-6 newtons. 25-50% above the upper limit of the specification (2-4N). The minibot itself struck the top-plate of the practice pole (wood plate) violently enough that the 10' steel pole actually jumped about 1/8". This would make sense if you do the physics of a 1.25 kg object traveling at about 2 m/s (some minibots are even faster than this) being stopped in less than 7mm of travel. It is a pretty violent impact. If the impact absorbed perfectly, I would expect a If you do the math on this, you would expect 7ms impact with a peak force around 30-40N.
Of course, this was week 1. I did hear that week 2 was improved immensely.

Assuming an elastic collision, this could have a dwell under 14ms. If 75ms is required, a flexible member requiring about 4N to flex it, and around 2-3 inches of compression just might work. Shorter if you are slower, longer if you are faster. For you guys going 3 m/s... Good luck. Hope to be joining you soon.

[johnr]

I haven't been to a competition yet, so i don't know, but are they letting you test your mini bot on the field before hand? Just to make sure it works. As far as the bolts go, if i was a ref i would put some chalk or something on the bolt heads. Something that would leave a mark. Or, before the comp the bots get checked to see if they can contact the bolt. Maybe a green mark on bot that can't touch bolt and a red mark if it can. This would end a part of the problem or questions that might arise.

[Ken Streeter]

Originally Posted by Brandon Holley

What bothers me is that there seemed to be some kind of a specification regarding force and now that specification seems less important than the time it takes to trigger the target.

Before I say anything else, let me start with a few caveats:

1. Our team (1519) was at a week zero (pre-ship) scrimmage (Nashua, NH) with the official field, and a week one tournament (NH).
2. At week zero, we had a minibot of around 3.5 pounds and 3-4 seconds.
3. At week one, we had a minibot of around 2.3 pounds and <2 seconds.
4. We did not compete at week two, so I don't have first-hand experience as to whether or not the field changes helped fix the problems.

At week zero, there were lots and lots of "false positives" occurring on the minibot towers from robots bumping into the base of the towers. My understanding is that this was due to very sensitive limit switches combined with an FMS sampling time of 20ms where only one sample was required to report a trigger. However, I don't recall any "false negatives" where a minibot climbed the pole, but didn't result in a successful trigger. Then again, I think there were only a couple minibots being successfully deployed at the scrimmage and these were still > 3 seconds in climb time.

At week one, in order to remedy the problem with false positives, the FMS software was changed (at least at GSR) so that a "trigger" would only occur when 8 consecutive positive samples were detected, with the samples still occuring at a 20ms rate. This FMS change effectively required minibots to hold at least one of the trigger limit switches closed for approximately 150ms. This change eliminated false positives from robots bumping into the base of the towers, and worked fine for slow minibots, but fast minibots were then having trouble hitting the trigger for LONG enough. On Thursday, our minibot was one of the ones which was regularly smashing into the top of the tower (with plenty of Newtons) but which was also not holding the switches closed long enough to result in a "trigger." Team 40 had a similarly quick minibot and was seeing the same problem.

We started adding a mechanical "impact absorber" to the top of our minibot in order to have it remain in contact with the switch plate a little longer, but this had two problems -- it didn't always extend the contact long enough, and sometimes failed to activate our off switch.

In the last match on Friday (a replay of Match 11), we unintentionally found a sure-fire solution to the problem of not contacting the tower for long enough (simply leave the motors on) as shown in the photo below:



(Separate photo discussion thread at http://www.chiefdelphi.com/forums/sh...ad.php?t=93430 )

However, our team budget isn't sufficient to afford cooking thermal protection wires and motors in every match, so we needed to come up with a different solution! We ended up increasing the reliability of our off-switch a little, but were still at the edge of triggering / not-triggering the FMS limit switches for long enough. Some matches it would work, and others it would not. I can only presume our contact time was very close to the required threshold.

In Week Two, it's widely known that FIRST made mechanical changes to the trigger assemblies. I'm curious as to whether or not they also changed the needed "contact time". (Maybe to the 75ms described earlier in this thread?) I do think that once a "contact time" is determined by FIRST, it would be helpful to release the information to teams.

That said, I'm still concerned that really fast minibots may not hold the trigger plate up for long enough to result in a positive trigger. It sounds to me that everything worked great at some regionals (Kettering and Pittsburgh have reported that) while other regionals may not have had as much success (San Diego?) This sounds to me that the problem may not yet be completely solved.

I'm curious to see what happens in Week Three regionals and hope that FIRST and teams continue to get this worked out. 1519 doesn't compete again until week Five.

I do know that with ever-faster minibots and deployment systems, anything other than automated scoring (except maybe official video replay) will have a hard time distinguishing between four minibots that all hit the top of the tower with between 9 and 8 seconds remaining in the match! At the Week One Granite State Regional, Finals-2 was determined by the minibot race, (see http://www.chiefdelphi.com/forums/sh...ad.php?t=93314 ) where all 4 minibots hit the top with between 8 and 6 seconds remaining on the clock. Not all of the minibots activated the trigger. The referees manually made the correct on-field call, but there was a fair bit of debate about whether or not they had. Counting on the referees to always be the adjudicators puts them in a really tough spot, so I hope that the automatic triggering system can be made to work reliably without having to pose major additional constraints on teams (such as requiring them to hold the trigger plate up for a long time.)

However, since there presumably is some sort of "contact time" required to have a successful trigger recorded, it is very important (as Brandon mentions) to have teams know what that time is in order to be able to adapt their minibots to satisfy the requirement.

[sanddrag]

Originally Posted by Brandon Holley

I think we can agree that the teams responsibility is to trigger the target, what we need is some direction from FIRST on what exactly that entails (your Q&A suggestion is a good one).

Originally Posted by Ken Streeter

At week one, in order to remedy the problem with false positives, the FMS software was changed (at least at GSR) so that a "trigger" would only occur when 8 consecutive positive samples were detected, with the samples still occuring at a 20ms rate. This FMS change effectively required minibots to hold at least one of the trigger limit switches closed for approximately 150ms. This change eliminated false positives from robots bumping into the base of the towers, and worked fine for slow minibots, but fast minibots were then having trouble hitting the trigger for LONG enough. On Thursday, our minibot was one of the ones which was regularly smashing into the top of the tower (with plenty of Newtons) but which was also not holding the switches closed long enough to result in a "trigger." Team 40 had a similarly quick minibot and was seeing the same problem.

This is the problem exactly. An engineering specification of what constitutes the act of TRIGGERing the tower has not been provided beyond that it takes 2-4 Newtons of force, and unspecified magic happens in the FMS to determine if it is TRIGGERed.

To the best of my knowledge, the manual or arena did not provide any specifcation as to how the tower switches are being sampled, or what they're connected to. I think it's fair enough to say we don't care about the speed of the electrons in the wires, but beyond that, whatever is happening post-switches could make a difference in the design, or in who wins.

For this sort of thing, 20ms is a ridiculously long sample time, and to require 8 consecutive samples at that slow of a rate is just absurd. If that's what it takes to prevent hostbot shakes from triggering it, it needs a redesign. Did FIRST seriously not think teams would build minibots that would bounce off the top in less than 1/10 second?

It was never specified for what duration the force must be applied.

I'd like to see a specification that says something like:

The TOWER is considered TRIGGERed when the switches, connected through _________ to a ________ controller, running code available for download from _____, cause the ____ variable to become true.

[Jared Russell]

In Florida, it appeared the vast majority of races resulted in the towers triggering properly, whether the minibot was slow or fast (and Florida had several sub-1.5 second minibots).

However, I do remember a couple of specific instances where the tower was not triggered for one reason or another (but not the teams involved - all I know is it wasn't us!).

[RMiller]

Originally Posted by jspatz1

Regarding "the bolts"....it seems that they should have been fixed to the lower plate, and moving through the upper plate, so that they move with the lower plate no matter where it is contacted, rather than being fixed to the top plate and protruding through the bottom plate, so that they create 4 areas which block the minibot from triggering. From my understanding of the drawings, this would be an easy change to make.

At Lake Superior, one of the FTAA thought of this solution as well. However, it was not implemented because it adds to the weight needed to push the plate up. I believe it was Bill Miller himself who said the FIRST folks thought of your idea, but realized the issue.

For the record, with the exception of issues across the entire field in the second finals match, I did not know of any issues with the towers at Lake Superior with the exception for one team that initially was hitting the bolt heads (they made a tiny modification to fix the issue). That said, the minibots were used by under half of the teams.

[Lil' Lavery]

Originally Posted by squirrel

I put more credence with those who read the rules and understand what the rules require the TEAMS to do.

There is no "should have TRIGGERED the TARGET" in the rules. You do it, you get credit. You don't do it, you don't get credit for it.

Part of the design challenge is to make sure your MINIBOT TRIGGERS the TOWER. Some TEAMS seem to have missed that. Apparently they thought the challenge was to get a minibot up the TOWER as quickly as possible.

By this logic, should the "false positives" from week 0 and thursday/friday in week 1 have counted so long as they occured in the last 10 seconds of a match? Should slamming your minibot into the tower hard enough to trigger the target count as winning the minibot race, even if your minibot never even leaves the hostbot?

[MagiChau]

Originally Posted by Lil' Lavery

By this logic, should the "false positives" from week 0 and thursday/friday in week 1 have counted so long as they occured in the last 10 seconds of a match? Should slamming your minibot into the tower hard enough to trigger the target count as winning the minibot race, even if your minibot never even leaves the hostbot?

Trigger is listed as the plate being pushed.

TRIGGERED – the act of pushing the bottom disk of the TARGET so that the sensors are tripped and a
signal is sent to the Field Management System (FMS). When a TARGET is TRIGGERED, the
MINIBOT RACE on that TOWER is complete.