Laser scan matching for measurement update in a particle filter

Abstract

The position estimation techniques based on particle filters use a motion model, to predict the possible robot poses, called particles, when a motion command is executed, and a measurement model to update the weights of these particles. Assigning weights to these particles on arrival of a new measurement is a fundamental problem. Different models mostly ad-hoc exist for this measurement update. Laser scanners are the most popular sensors used for sensing the environment during navigation. We demonstrate a method to assign weights to position predictions by matching two laser scans. This method is based on an Euclidean metric. We believe that it is sensible to use an Euclidean metric to find the degree of match between two laser scans in an x, y plane. The results of measurement update using this method are presented in real-time simulated environment.