Self-localization by using artificial neural networks for humanoid robot NAO

NAO is a humanoid robot that is widely used in robot soccer games. Position estimation is an important process in such robotics applications. It can be defined as finding the position of the robot in a known environment. The current solutions proposed in the literature usually utilize proximity markers that provide the necessary information to determine the position of the robot. However, self-localization without using an external marker is a challenging problem. In this study, a novel approach that is based on Artificial Neural Network (ANN) learning is proposed for the self-localization problem of robot NAO on a soccer field. The method uses images captured by the vision sensors of the robot and a supervised learning process is carried out in order to obtain a self localization system. Some image processing methods are also utilized in order to extract the features that are used in the learning process. Various tests are carried out and it has been observed that the NAO robot can estimate its position on the soccer field quite accurately.