Representation of odometry errors on occupancy grids

Abstract

In this work we propose an enhanced model for mapping from sonar sensors and odometry that allows a robot to represent an environment map in a more suitable way to both the sonar sensory data and odometry system of the robot. We use a stochastic modelling of the errors that brings up reliable information. As a contribution, we obtain a final map that is more coherent with the reality of the original data provided by the robotic system. Practical experiments show the results obtained with the proposed modification to be trustable in such a way that this map can be used to provide previous knowledge to the mobile robot in order to perform its tasks in an easier and accurate way. Moreover, the map can help the robot to support unexpected situations inside of the environment.