The core of the wrist’s stability comes from an intense focus on software validation. Before we ran any significant code changes at a competition, we had already run dozens of practice cycles at home to identify and resolve as many problems as possible.
This also meant that we gradually stepped up our software integration as the season went on, slowly adding more functions as they were extensively validated. At our first competition we ran the wrist under 100% manual control, at our second the wrist was limited by limit switches, at the third it was PID stabilized, etc.
This gradual step up eventually ended where we were at worlds, with PID-controlled preset angles for numerous actions, limit-switch soft stops to prevent self destruction, and a 100% manual override if anything was to fail.
Another huge part of the wrist’s stability is pre-match function checks during competitions. We are very thorough and test every wrist and elevator preset available to the drivers. If we had missed a limit switch being broken or the encoder counting incorrectly, we’d notice that something is functioning incorrectly and either pursue a solution to the problem or decide to go to manual control.
Let us know if you have more specific questions. I also think we’ll have our code publicly released shortly!