This repository contains the central AI controller software framework used by Maracatronics Robotics Team and Research Lab at Small Size League
This is our release to Robocup Open - CBRobotica 2020 competition that contains new updates that are listed below:
-
Dynamic defenders
- Now our defenders are capable of more dynamic decision making on the field, mainly involving the way they mark enemies and the ball.
-
Multilayer perceptron on attackers
- With the use of this statistical learning algorithm, we make our attackers more solid and capable of more dynamic decision making involving passes and shots on goal, ensuring not only more possession of the ball but successful finalizations at goal.
-
Attackers positioning improved
- Now our attackers are positioning themselves in more strategic ways on the pitch, optimizing positions where they have the greatest openings for goal. This allows that, together with the decision layer for kicks and passes, we are able to get positions for passes (and consequently submissions) close to the optimum.
-
Optimizations and memory leak fixes
- Our old release contained serious problems: very high processing costs and serious memory leak problems with some parts of the software. For this release we tried to optimize all these factors to the maximum, and now we have software that is consuming more than half of the consumption of the old version.
- Unfortunately we have not yet managed, despite all the research, to improve the timeroverextended problems, but we are modeling a second version of the software that will avoid these problems among distributed systems.
Please, refer to Armorial-WS and install the entire Workspace.
To run the Armorial-Carrero please follow the steps listed in Armorial-WS.