Driving-Condition dependent and Monte Carlo Simulation-based Optimization Method for a Bayesian Localization Filter for Parking
In this paper, a method for the automated adaptive tuning of a Bayesian localization filter is investigated, since previously parameters were often only empirically determined. The method determines the optimal parameters offline using Monte Carlo Simulation (MCS) results and ground truth data. The method is applied to a previously developed Bayesian localization filter called Odometry 2.0 estimator. Its architecture makes it possible to merge different dead-reckoning models individually, since the input is redundant. However, under different driving-conditions (DCs) during parking, the individual models can have advantages and disadvantages. To create a DC dependent optimization, different cluster analysis methods are investigated to automatically divide a parking maneuver into different useful segments. Finally, these segments and statistical error and coefficient models are used in the MCS. The error model provides sensor errors for robust tuning and the coefficient model provides randomly found parameters. The results of the optimization show an increase of the robustness and are again significantly increased under consideration of the individual DCs, since effects that cannot be modelled are corrected.