The effect of changes in the environment on the performance of a SLAM algorithm

Abstract

In this thesis the effects of three variables on the results of a SLAM algorithm were analyzed. The manipulated variables are: added noise to the distance and motion sensor and slipperiness of the wheels. The aim of this thesis is to find out how these variables affect SLAM maps and if any structure can be found in the effects. This thesis proposes that a mapping can be learned by a Neural Network that would optimize SLAM maps by removing the effects. The research found that there is an effect on the maps for added noise to the distance and motion sensor of a robot. A structure in the effects was only identified for the noisy distance sensor variable.