Particle filter as the localization of the robot on the football pitch

Abstract

RoboCup competition is a football tournament played by robots. In one of the categories, matches are played between teams of humanoid robots. Accurate position tracking for each team member is essential for building more advanced strategies. However, rules of the game forbid using any sensors that allow global positioning. This article describes an implementation of a positioning system based on a particle system, which is a standard approach for localization problems in known environments. Proposed algorithm uses a monocular camera to obtain feedback from the environment and calculates the similarity between real observation and each particle. Developed system searches for the region with highest local density of particles for best state approximation.