A new global localization algorithm based on feature extraction and particle filter

This paper describes a new global localization algorithm based on feature extraction and particle filter. This algorithm uses two kinds of sensors: wheels encoders and a laser scanner. A map of the environment is built by taking laser readings of the environment from well-known poses of the robot. The resulting map is composed by a list of features, representing the position of clusters obtained by using the mean shift algorithm. The mean shift algorithm is also applied for each sampling step in order to calculate the importance factor of the particles. The trials have been conducted by using a simulator of a differential drive robot