Vision-Based Robot Position Estimation

Abstract

The purpose of this work is to determine whether it is possible or not to estimate the position of a robot on a soccer field thanks to a unique camera. It is realized in the context of the RoboCup contest: a competition where two teams of robots play against each other in a soccer game.

This paper aims to describe the algorithm developed to solve the positioning problem, and analyses the results obtained while testing on a prototype of the soccer field. The principle of the algorithm is first to use already existing image processing tools to extract specific information (such as the field edges and corners) from the images, and then, use those information as a basis for the position estimation.

The result analysis shows that the developed method allows to estimate the robot position with a mean error of 5 cm. This error does not increase with time since the position estimation is applied on every image independently. Therefore, this algorithm would be helpful in working with a more precise positioning method that sometimes needs to be recalibrated (such as inertial systems).