System bug need to be taking care

This is insane red did a great job scoring high goal of 41 cargos in teleop the system counted 67!!

FIRRST HQ mast gets us some answers to how they are going to deal with this bug during the CMP it can damage competitions,

5 Likes

No system is ever perfect but somebody definitely should have noticed when it jumped from 40 to 80. Was this match replayed? Obvious field fault. Kind of surprised the game announcer didn’t notice. It seems like it should be something as easy as moving the sensor location to farther down the exit ramps. Also many instances in other videos of cargo not being counted. This needs to be fixed!

1 Like

The match was not replayed. It was announced there were no scoring issues and the match would stand.

Without knowing anything about the problem,sensors used or architecture… I’d like to suggest 2 common solutions to similar problems:

  1. don’t accept new balls from the specific sensor for 100 mSec after that sensor reported a ball detected.

  2. implement sort of hysteresis thresholding - after reaching value X , require going below a lower value to declare “ball detected”

snoman

Well that is Crap, and a fail of the system. From 1:05 to 1:09 the red alliance scored 40pts in the hub, 20 cargo… in 4 seconds. I think not. Would it have changed the outcome? Probably not in this match. If I was a student on the blue alliance, I would have had to been physically removed from the question box. This wasn’t a missed or questionable call by a ref.

Having been behind the scoring table, I can say that there are cases where the field itself reports scoring issues. There are times where one of the sensors may report a negative number of balls, or where the sensors don’t agree - for example, it detects 3 balls, but only 1 red and 1 blue. These situations pop up on the scoring computer as an error that has to be dealt with. In many cases, a difference of 1 ball either way isn’t going to change the outcome, so they assign it one way or the other and move on. In others, it may have a significant impact and require the match to be replayed.

In this case, the volunteers on the field have no way to verify there was an issue. Presuming the system didn’t identify something and alert them, they’re left with the system telling them one thing and teams telling them another - and remember, with no video review allowed, it’s impossible for them (assuming they didn’t notice it as it happened) to really determine what happened.

From watching the video and comparing to the official numbers on frc-events:

  • In auto, two red balls go through the upper near port, but only 1 is scored.
  • In teleop, 8 red balls were scored in the upper near goal, versus 35 reported. Otherwise, I come up with 1 other difference - 15 red balls scored in the upper far goal, versus 14 reported.

At least in this case, we seem to be able to narrow it down to a single set of sensors on the upper near goal. Of the 27 that it scored extra, about 20 of those all came at the same time, with 92 seconds left in the match. The rest appear to happen with 62 seconds left. In each case, there is a red ball passing through the upper near goal. I can’t be sure, but looking at the video a few times, it might be the same ball in both cases (the same ball is definitely in the hub the second time, I’m just not sure it goes through the same port, it’s hard to tell). I wonder if there is some ball condition (excessive fluffiness, or a bald spot, or something like that) that can cause this.

3 Likes

You don’t need video review. Many teams have scouting data that shows the count within 1 or 2 cargo. No prohibition on using that.

5 Likes

I would say that “team scouting data” would land under the “etc.” of that particular rule. (And while I’d trust that teams as a whole are making their best effort to get good and accurate data, I would not say that I’d trust that specific data being presented to be 100% accurate.)

Now, the HR is allowed to get input from other sources, but all of those listed sources are FIRST-trained (or FIRST-trained by proxy).

Don’t get me wrong: The fact this bug still exists is really bad. And I think it should be grounds for a replay if it’s >1 or 2 cargo in any given direction. And there SHOULD be corrective measures (manual scoring, most likely).

1 Like

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.