Effect of Resampling Methods to Performance of FastSLAM Under Different Noise Conditions

In this study, variation of estimation errors of resampling methods which is one of the most important steps of FastSLAM algorithm, in different process and measurement noise values under different particle numbers is examined. It is seen that variation of process noise affected error values more than variation of measurement noise for all resampling methods, and Metropolis resampling is the method least affected by variation of measurement noise. It has been determined that resampling method that provides the closest error value to the correct position changes according to the noise conditions in which the system operates.