Decision tree analysis for humanoid robot soccer goalkeeper algorithm

Abstract

Latest humanoid robot soccer competition rule specifies larger field, wider goal line, and more number of robotics available for a team than before. In consequence, the humanoid goalkeeper needs to be more responsive in adjusting its position to follow ball movement and to react precisely to block incoming shots. Both the robotic control system and algorithm have to be refined. In this article, the humanoid goalkeeper algorithm is extended by utilizing compass to help with direction, using inter-robots communication to share information regarding ball location, using head servo orientation to look around, providing vision via Android smartphone camera, and employ landmarks in the field to provide guidance. There are many attributes perceived by robot sensors in preparing robotic motions and strategy. Each attribute is considered as a random event with uncertain values. Here, decision tree analysis based on entropy and information gain are used to evaluate the uncertainty. Overall, there are some improvements for the humanoid goalkeeper performance.