Sistema de posicionamento de robôs em partidas de futebol baseado em inteligência coletiva por enxame

The RoboFEI Small Size League (SSL) team was created in 2008. One of the motivations for the project is the application of knowledge in electronics, mechanics, and programming in the development of algorithms focused on Artificial Intelligence (AI). AI covers several techniques, such as learning, optimization, and bioinspired algorithms. Bioinspired algorithms are used for the most diverse purposes, including so that robots can work collaboratively. The SSL has evolved over the years. Some changes have already been made, such as increasing the size of the field and the number of robots. This evolution also brings greater possibilities for moves and increased complexity of a match. The positioning of robots in the field becomes essential as a defense and attack mechanism. In this scenario, this work proposes the use of the Particle Swarm Optimization (PSO) algorithm as a collective intelligence option applied to determine the positioning of robots in soccer matches. Fitness functions were proposed for defending the goal and blocking passes in the SSL. In order to develop these functions, tactical principles of modern soccer games were verified. To assess the effectiveness of the optimization, functions metrics proposed to measure the Positioning Performance Index (PPI) of the original and optimized positions. These metrics based on the System of Tactical Assessment in Soccer (FUT-SAT) that defines the Tactical Performance Index (TPI) of a team based on specific criteria and positions in the field. To assess the effectiveness of the fitness functions, plays with effective goals from RoboCup 2019 – A league were selected. These plays were separated the match, from the beginning of the touch of the ball until the goal finish, at intervals of 200 milliseconds and called as “instants”. For each instant, the positioning of the defense is optimized. In the end, the evaluation metrics of the new positioning are applied and compared with the original. The evaluation and visual inspection metrics show that the suggested positions could have prevented the play’s continuation at several moments before the goal of the adversary team. The experiments demonstrated the effectiveness of optimization and metrics. Finally, we can apply fitness and metrics functions in other categories of soccer robots