Reducing the error in mobile location estimation using robust window functions

The non-symmetric nature of the distortion that NLoS propagation introduces into the measurements when performing location estimation from distance measurements makes standard robust statistical estimation techniques relatively ineffective. In this paper two new window functions obtained by modifying the conventional Huber and Tukey window functions are presented to reduce the error in distance measurements that are corrupted by multipath fading. It was found that for the Gaussian distributed NLoS component present in our data, both the modified Huber and modified Tukey window functions yield lower root mean square error (RMSE) compared to the maximum likelihood estimator and the original Huber window function. Results also show that the modified window functions are more robust to the data outliers created by NLoS measurements than conventional robust estimation algorithms.