An analysis of Reinforcement Learning applied to Coach task in IEEE Very Small Size Soccer

The IEEE Very Small Size Soccer (VSSS) is a robot soccer competition in which\ntwo teams of three small robots play against each other. Traditionally, a\ndeterministic coach agent will choose the most suitable strategy and formation\nfor each adversary's strategy. Therefore, the role of a coach is of great\nimportance to the game. In this sense, this paper proposes an end-to-end\napproach for the coaching task based on Reinforcement Learning (RL). The\nproposed system processes the information during the simulated matches to learn\nan optimal policy that chooses the current formation, depending on the opponent\nand game conditions. We trained two RL policies against three different teams\n(balanced, offensive, and heavily offensive) in a simulated environment. Our\nresults were assessed against one of the top teams of the VSSS league, showing\npromising results after achieving a win/loss ratio of approximately 2.0.\n