Hey,
I’m struggling to understand the rules of the Galactic Search Challenge.
“…by locating and collecting Power Cells as fast as they can on one of two pairs of paths”
“The raw score for this challenge is the sum of the completion times (in sec) for two paths runs…”
Do I need to submit one of each path (either red or blue randomly)? or, can I choose one of the 2 paths and submit one of them?
THX, Ofek Harel. BumbleB 3339.
At the beginning of an attempt, you will flip a coin to determine if the power cells will be set up on the red path or the blue path. Without communicating with the robot, then you run an autonomous that’s capable of picking up all 3 balls and going from the start zone to the end zone. There are 2 “paths”, path a and b, and two colors, red and blue. One complete submission is either path a and b using only the red path, or path a and b using only the blue path.
the randomization aspect seems really easy to circumnavigate. Lets say your team makes an autonomous program that only works if you randomly get the red paths. If you then randomly get the blue paths, there is literally nothing obligating you to do the blue paths. Unless I’m reading the rules wrong, I’m pretty sure that you could just scrap that attempt and start all the way from the top. Sure, about 50% of your runs would “fail”, but all you have to do is get 1 good run and submit it, you don’t actually have to worry about being consistent. I can only assume that the randomization aspect of this challenge is to force the autonomous programs to be more adaptive, but i cant think of anyway that FIRST could really fix this loophole without fundamentally changing how the randomization process worked.
I did some tests yesterday. Here’s the video
My team was talking about this too. It seems a little too easy, but we couldn’t find any rule against it or even really discouraging it. I guess they could fix it by requiring you to stream video live but that would be a pretty dramatic change and adding “with very good wifi” to the 15’x30’ space requirement would be pretty burdensome.
They could change it to require you to do submit a video of you doing both colors…
You could have your robot also randomly choose red vs blue - it would only get it right 1/4 of the time, but would give you video of both colors if you ran it enough times. Maybe they could fix it by requiring you to submit a video of your robot doing the challenge 10 times (for example) in one long take. But frankly, updating the rules from “consistency not required in any way, shape, or form” to “will be evaluated on consistency” would be a pretty massive change.
This is all hypothetical - a year where there’s not a lot else going on is a good time to try and learn how to incorporate vision processing into our auton, and I expect that’s what our programmers will want to do. But still, the rules as written seem easy to game.
Agreed, you can just have it guess and selectively submit, but it does have the added benefit of requiring teams to make autos for both routes, which isn’t necessary RAW.
It seems like the best way to fix this would be to use just 1 of the paths and say that teams must do both red and blue with the same code (honor thing as there is no way to check this) so that it has to at least involve something that makes a decision about which path to run.
Couldn’t you just make the program alternate between red and blue?
Are we 100% sure that the robot won’t know which path is chosen? I just remember the 2018 game, where one was randomly assigned a side of the scale if I recall correctly. Even though the drivers were not informed ahead of time, in code, you were given which side you could place the cubes in as part of the context.
My interpretation of the rules is that the robot doesn’t know what set of balls is chosen until autonomous is enabled. And it has to determine that on its own, not from human input. Or it just ignores the random chance and follows a path that would hit all 6 positions.
Any more knowledge than that (like setting it down but then a human picks between 2 auto paths on the driver-station before hitting start) gets into a shady area that I argue is disingenuous.
EDIT: My example actually is past shady, and into straight rule-breaking. In my reading of the rules.
Oh no–I was referring to the drivers being completely unaware of the result but the program having the context.
You would only write one auto program, but you can call something like getFieldContext() or something along those lines in your single auto program and then do different things depending on what it returns (e.g., if getFieldContext() returns red, follow the red path).
So yes, the robot wouldn’t know which set of balls is chosen until autonomous is enabled, at which point it would from field context. I don’t know if they’ll provide that functionality this year, but I believe they have in past years.
Oh sure. You’re correct that the Field Management System (FMS) provided that info in 2018. But there is no FMS system running the At Home portion of this game, so I am 99% certain there won’t be a method you could call to get that info.
Ah, got it. Thanks for clarifying–I wasn’t aware that there wasn’t a FMS this year. I suppose manually entering the result of your coin flip to the computer, which is how it would have to be done, is tantamount to just selecting one of two auto programs.
I think everyone agrees that the robot and the autonomous choice (which auto to run) should be before the color is chosen.
What everyone here is questioning is if a team would need something (like vision) that actively decides where the balls are at. I think the intent of the rules are for teams to sense this and adapt on the fly. However I think it is completely within the rules as written to just write an auto assuming a color (no decision) and then discard any runs when the other color is randomly picked.
That is the loophole that we are trying to close assuming FIRST actually wants to close it.
Yeah was just trying to answer his question since no one had responded yet.
I think FIRST’s intention is similar to 2018, like nIspector said. And that year, I know some teams just hoped that the colors lined up appropriately for where they lined up their robot. The biggest difference between now and 2018 is you have an infinite number of matches and only have to count the ones you want to count.
I agree with a previous poster that you should be running the same code each time though. Even if the code just picks randomly between Red and Blue.
You could get getGameSpecificMessage() but that just reflects a text box in the DS, which is likely considered signaling.
And I would consider that “the team signaling to the robot”. Which is what is expressly not allowed.
Exactly. At least if a team member is typing it in
DISCLAIMER DON’T DO THIS
If the person typing it in isn’t registered on the team, then the team does not signal to the ROBOT.