Real-Time Scoring Not Working?

[MooreteP]

Originally Posted by FrankJ

Load cells used in weighing systems are generally accurate & robust. The problems come up with how they are placed into the system. The goals need to be total supported by the cells with no contact with anything else. I aspect this will get better as the weeks go on & lessons learned get incorporated into field set up.

They are totally supported by the cells.

[Chris Fultz]

i was a field volunteer at BAE.

From the start, we were counting discs as a verification of the score. When the counts were below about 15, it was easy to count from behind the player stations. When the counts were higher, or we were not confident, then we did a manual unload and count and the scores were not input until we submitted the human counted counts. We counted after Auton and at the end of the match.

For elims, there were two people at each end just counting and we had to agree.

All of the volunteers recognized the importance of accuracy, and how one disc can make an impact. I have a lot of confidence in the BAE scores.

I think FIRST will create an improved system, but it may take more than one week since the trucks are already on their way to week 2 events.

[CalTran]

Originally Posted by Chris Fultz

We counted after Auton and at the end of the match.

Chris, what's the protocol for verifying an Autonomous score? Can you elaborate on the "snapshot" concept?

[Undertones]

After Einstein last year, I'm sure FIRST will be taking every measure to make sure the field works as designed.

[Ian Curtis]

Originally Posted by FrankJ

Load cells used in weighing systems are generally accurate & robust. The problems come up with how they are placed into the system. The goals need to be total supported by the cells with no contact with anything else. I aspect this will get better as the weeks go on & lessons learned get incorporated into field set up.

In college, my Design/Build/Fly team tried to build a wind tunnel balance. We were unsuccessful at convincing any of the faculty let us use a real load cell, so we found one on the internet that was surprisingly cheap. We bought it, and implemented it in our design. Big surprise, it did not work at all. In back to back runs it would give us totally different answers. We ended up using a postal scale, which worked great for our purposes! Another guy on that team checked out the HH field, and apparently FIRST is using that same load cell.

[Chris Fultz]

Originally Posted by CalTran

Chris, what's the protocol for verifying an Autonomous score? Can you elaborate on the "snapshot" concept?

We counted and recorded the discs after auton was over. Those counts were easy because the most you could have was 9, and it was easy to count them coming in, and from looking up into the box. The person at the field table also counted as they went in (there was a team of people for red and for blue alliance ends).

At the end of the match, we then just did a total count. The person recording the scores (on paper) then did the math to write in the number scored in teleop. This written score was then handed to the official scorekeeper to input into the system.

[Greg McKaskle]

and apparently FIRST is using that same load cell

This seems a little vague. Can you give more details? What internet sensor were you using, what circuit was the sensor in, what was the resolution and rate of the measurement?

Greg McKaskle

[Ian Curtis]

Originally Posted by Greg McKaskle

This seems a little vague. Can you give more details? What internet sensor were you using, what circuit was the sensor in, what was the resolution and rate of the measurement?

Greg McKaskle

Probably because it is a little vague, and perhaps a little premature. I believe was a custom Arduino rig, but I am fuzzy on the details. Basically we took raw voltmeter readings -- clearly you could do a lot post processing to smooth it out. I've attached one of the resulting data files -- as I recall our issue was not with a given run giving weird data, it is that you just couldn't get the same thing twice back to back. Clearly we were sampling way faster than the sensor was updating. We were using it as a thrust test stand, which is not a great environment for a load cell because there is a bunch of vibration. However, we had similar issues even using bricks. Using an electronic postal scale to measure the force at the same location on the lever, these issues went away.

I don't think I have any of the documentation on my computer anymore, but I'll email around and see if someone else has the part number.

EDIT: FC-22, the 0-10 lb version. I can't find somewhere where FIRST lists the sensor that they used, but this family of sensors pops up quite a big when I google around for low cost load cells. Anyone "in the know" or take a look at the field to confirm?

[cgmv123]

Originally Posted by Chris Fultz

I think FIRST will create an improved system, but it may take more than one week since the trucks are already on their way to week 2 events.

Hopefully it can just be a software update to further refine and process the inputs, which can be applied at any time. As noted, weight counting is effective when used properly.

[Grim Tuesday]

Originally Posted by Chris Fultz

We counted and recorded the discs after auton was over. Those counts were easy because the most you could have was 9, and it was easy to count them coming in, and from looking up into the box. The person at the field table also counted as they went in (there was a team of people for red and for blue alliance ends).

At the end of the match, we then just did a total count. The person recording the scores (on paper) then did the math to write in the number scored in teleop. This written score was then handed to the official scorekeeper to input into the system.

It seems like this is one of the few cases where high speed cameras might be useful. What did you guys do for goals that happened right at the buzzer?

[Lil' Lavery]

Originally Posted by DonRotolo

Good choice of words Gary: Unfortunate. I am sure everyone had hoped it would work, but that's just not how it works in the real world.

I can't do better; anyone who can should do so before the end of the season and give one to FIRST for their use. Hey, maybe you'll end up on the GDC...

I've got a better word than unfortunate. Unacceptable.
This is far from the first time that FIRST has screwed up real time scoring. It's a complete joke at this point. To has real time scores as far off as these are completely erodes the public and team's trust in FIRST and the GDC. Between this and the field delays, I actually had a student ask me something along the lines of "how can such smart people screw up so badly?" This isn't something that we should just shrug off. Just about everyone saw this coming from miles away, so why FIRST thought it would be an acceptable solution is beyond me.

And it goes beyond the load sensors. As someone already pointed out, the FMS is displaying odd scored in autonomous. Odd scores aren't even possible in autonomous. "61" was a frequent score for the blue alliance during autonomous at Hatboro-Horsham. You can score 60 points in auto. You can score 62 points in auto. You cannot score 61.

As for a better option, humans. That's what they're using now to get accurate scores, might as well use it for real-time as well. Aside of 2010 and 2012, the only accurate real time scoring FRC has ever used has been input by humans. 2004 was humans. 2005 was humans. 2007 was humans. 2008 was humans. 2011 was humans. The only time human RTS has been lacking was in 2003 and 2009, but automated scoring wouldn't have faired any better in those chaotic games.

Originally Posted by Chris Fultz

We counted and recorded the discs after auton was over. Those counts were easy because the most you could have was 9, and it was easy to count them coming in, and from looking up into the box. The person at the field table also counted as they went in (there was a team of people for red and for blue alliance ends).

The max in 9? Tell that to 365, 103, and 2559. They scored 11 in SF2-2 at Hatboro-Horsham.

[mathking]

I really wish the GDC would have anticipated the potential problems. I can't imagine, given how widespread they were, that this didn't crop up before. As Sean and some others have pointed out, humans could do the real time scoring. I think that pads like they had in 2011 would work for a good approximation for the RTS. Add in a manual count for verification at the end of each match and I think you would have a perfectly acceptable alternative.

[Jaxom]

Originally Posted by Chris Fultz

Those counts were easy because the most you could have was 9.

Really? Our alliance scored 62 in auto in one of the finals in Hub City, and I'm pretty sure that wasn't the high auto score for week one. I'm not sure, but I think that was 7 6-pointers from us, 2 6-pointers from 3931, and 2 4-pointers from 1801. Still easy to count, but it might mean someone has to take off their shoes.

[Lil' Lavery]

Originally Posted by Jaxom

Really? Our alliance scored 62 in auto in one of the finals in Hub City, and I'm pretty sure that wasn't the high auto score for week one. I'm not sure, but I think that was 7 6-pointers from us, 2 6-pointers from 3931, and 2 4-pointers from 1801. Still easy to count, but it might mean someone has to take off their shoes.

The highest I know of was either 66 or 72 points by 365, 103, and 2559. I can't remember if they got one autonomous match to work perfectly or not (I know they definitely had a match where they hit 11 shots and one fell out of 103 before shooting). 2559 would score 2 from the front right corner of the pyramid. 103 would score 3 from the back right corner, then drive underneath the pyramid to collect 2, and score those. 365 would score three from the rear center of the pyramid, drive backward to collect 2 from the centerline and score those. All of them at the three point goal. Both 103 and 365 would occasionally have some issues collecting their additional disks, but pretty much nobody on that alliance would miss. It was incredible.

[Ivan Malik]

While I agree that at this point human counting is probably the quickest, easiest and, by far, the cheapest solution to the inaccurate scores. How exactly are they going to be implemented into the existing field set up for a real time score? Not only software and hardware, but where would you put them? They can't be behind the goals; you can't see most of the discs enter up top or the pyramid. They cant be at mid field; you cant see the 1pt goals. It is also extremely hard to see the pyramid goals from field side. You would need at least 3 pads per alliance to even have a shot at being right.

When you add in the speed that discs will be scored during the later regionals/districts, let alone MSC, MAR champs, and St.Louis, and there is no way you will get an accurate real time score from a real time human count.

This game is simply not designed for human real time scoring. 100 some odd game pieces flying in multiple directions, being scored in 8 hard to see containers... I sure as heck don't want to be the one with that job. Its a similar situation to 2011, and that worked oh so well.