RedUtilities works using the RLBot framework, which is a framework for making and playing with these bots. The RLBot framework also has direct support from Psyonix, the creators of Rocket League.

RedUtilities is loosly based on another set of RLBot utilities made in python, called GoslingUtils. In GoslingUtils, bots would push "Routines" to the routine stack, and then the top routine would run every frame until it was completed, then the routine would be popped from the stack, and the next routine would run. In RedUtilites, bots run "Actions", only one action could be running at a time, and there isn't any stack. Instead, actions often had sub-actions running inside, and possibly sub-actions within sub-actions.

RedUtilities also comes with many actions built in, like aerial actions, double jump actions, and so on. These actions used heavy amounts of physics and linear algebra to position the car correctly during the action.

The example bot for RedUtilities, RedBot, managed to get 3rd in the 2022 RLBot Championship. It was only beaten by two machine learning bots, making RedBot the best non-maching learning bot in the entire tournament. You can check out RedUtilities on github, or you can get started using RedUtilities by following my tutorial series.

---