Hi:)
As we work on developing our scouting and strategy systems for the upcoming season, I’d love to hear your insights!
What are the features you’ve always wanted in a scouting or strategy system? Or, if you’ve developed your own system, what game-changing features have you included?
If you haven’t created a system yourself, what features in the systems you’ve used have stood out as particularly useful or impactful?
Thanks!
The biggest game-changing feature I’ve seen is Lovat’s path viewer. A nice feature i like is included match scouter scheduling; it prevents two scouters from watching the same robot.
So if we really are going for DREAM system boy do I have one.
A calibrated camera set up in the stands that uses AI tracking to understand the relative position of each robot on the field.
Humans then just have a button of “intake” or “shoot”, perhaps misses? and they press each button when that thing happens (of course this would be tailored according to the game)
Then based on when the button was pressed, that data would be matched up with the position data. Now all of a sudden you have full tracking of auto routines, accuracy of shots from different positions, places they intake from, cycle times and more. Essentially its the ultimate data hoarding technique, that you can then pull almost any data point from.
Similar (but different i think) thread here :
If you can track robots, you can track game pieces - at least every game piece is identical. So you wouldn’t need humans.
We experimented with time stamping all activites tracked in our scounting app. This would provide data necessary to get firm cycle times and such.
In '24 we were specifically focused on when teams reached the center line game pieces in auto so we could respond accordingly with our auto routine. The concept was sound. Did it work? We won the automonous award on the Newton field and scored 51 in auto first match of elims against 254. In practice, our operator prescouted every team at all our events to know our opponents tendancies in auto so we could gain a material advantage in auto. For worlds, he pre-scouted 75 teams.
This year, something more automated would be ideal.
That’s really cool. I think I came across a tool on a CD thread that might do something similar.
The challenge comes when you need to position a camera in a location where you have to ensure people don’t pass through, as relying on the streams can be unreliable - at least in my area.
Technically, it’s a bit complex but definitely doable. I remember seeing a tool that accepts an endpoint and generates robot trajectories based on it, but I haven’t been able to find it again.
I’ve been working on our own scouting app for the last few months, and here are the killer features that I am really proud of so far:
My ideal would be on the backend — take the data from quals and then run Monte Carlo simulation on every possible set of alliances. Then as selections are made all the alliances made impossible by the selection are dropped. When it is our turn to choose we should see the best remaining alliances possible for us in front of us. An additional refinement would be to use the data to predict the optimal choices for other teams to mark those alliances as unlikely to be available by the time our 2nd pick comes up.
Our scouting system has a match planning tool built into it. All the teams playing in your next match are listed along with some key stats to inform your strategy. There is a whiteboard of the field that you can draw on. Positional stats recorded by the scouts appear on it (and can be removed with check boxes).
You can use this link to play with the actual product.
Scouting, Strategy and Drive team should all be on the same page. The biggest issue with scouting is keeping students interest up. Ideally you want to figure out what each team does. How a team approaches that goal of maintaining interest (= quality Intel) , interest= buy in, and knowlege that can be used is a fine balance.
