Score Optimization with Linear Programming

#1

Score Optimization Presentation .pdf (267.9 KB)

Using scouting data collected, the Excel program can find the best combination of which alliance partner should score hatches and cargos at what locations in order to maximize score.

9 Likes

#2

Does this account for pathing conflicts? With 3 robots and only 2 independent routes it is important to understand the pathing of every robot to avoid as many intersections as possible. The scouting data may get too complex to accurately optimize this aspect.

0 Likes

#3

Thank you for sharing this, Ed. Math is awesome. This is awesome :slight_smile:

I have a few questions.

  • What scouting data is incorporated into this approach? Is it cycle times, scored game piece counts, or something else?
  • How are these numbers formulated into constraints?
  • Is any attempt made to account for the variability of team performance, or is it assumed that each team will play at their average level?
  • How much was this used for planning actual match strategy, and how effective was it?
  • Was there any discussion of incorporating defense into this model?

2791’s scouts and I discussed some similar approaches to modeling optimal scoring this year but did not follow through on any of them. I’m glad to see someone did though!

0 Likes

#4

No this is to maximize score only. It does not take into account of the path. There are so many variables and so many things that can happen in a match. Planning every second of the match is futile. Alliance partners can figure out the best way not to get into each other’s way.

0 Likes

#5
  • What scouting data is incorporated into this approach? Is it cycle times, scored game piece counts, or something else?
    To do this calculation, you need (1) max number of hatches scored in a match out of all previous matches at each level, (2) max number of cargo scored in a match out of all previous matches at each level, (3) max number of game pieces (hatch or cargo) scored in a match out of all previous matches.
  • How are these numbers formulated into constraints?
    The constraints are the solution can not exceed a team’s capability in scoring hatches, cargo and total number of game pieces in a match.
  • Is any attempt made to account for the variability of team performance, or is it assumed that each team will play at their average level?
    This is not to predict how a team will probably perform. This is a calculation of the maximum score an alliance can do with no defense.
  • How much was this used for planning actual match strategy, and how effective was it?
    I was with the team at Del Mar. We used it before every match in conjunction with visual data analysis from Team 8’s scouting system. It definitely helped. I was not with the team at Great Northern. They used it when needed.
  • Was there any discussion of incorporating defense into this model?
    The results helped us decide if defense or counter-defense is helpful. That decision can be automated also. Right now we decided to leave something for the humans to do.
5 Likes

#6

Team 8 will be at Silicon Valley Regional next weekend. If you would like to see how the spreadsheet works with live data, stop by their pit. Ask for Caleb, Bryan or Robbie. They will be happy to show you.

0 Likes

#7

will the team be making the excel file available at anytime to see the model in action for those of us nowhere near Cali? It is definitely an interesting idea that I was thinking of but never had the time to dive deeper into.

0 Likes

#8

We plan to test it again at Silicon Valley regional this weekend. If we are happy everything is working as intended, we will make plans to release the spreadsheet after the event.

0 Likes