Leveraging RSS Data For An Improved Radiolocation Estimation Algorithm Realization In LoRaWAN Using A Two-Tier Normalization

Abstract

An attempt to further enhance the accuracy and reliability of M-iWMLR localization algorithm using a new weight matrix that was formulated with two-tier RSS data normalization is presented. The two-tier normalization: data clipping and z-score normalization were applied to form a new weight matrix in this work. Data clipping was first applied to reduce significantly the effects of outliers on the RSS data while z-score normalization provides data consistency. The new localization algorithm herein, referred to as Ext.M-iWMLR algorithm is carefully evaluated by the use of location accuracy (location error), root mean square error (RMSE), range of error, and R2 score metrics. This algorithm is validated with the Modified Improved Weighted Multiple Linear Regression (M-iWMLR). The simulation results generated with MATLAB show that the Ext.M-iWMLR algorithm, at 95 percentile reduced the mean location error by 19.45%. The range of error and RMSE are reduced by 11.08% and 17.95%, respectively. Furthermore, the respective R2 scores were increased by 5.71% and 17.17% for the latitude and longitude coordinates. It was established that the new weight matrix formulated through two-step normalization enhanced all the considered metrics.

Keywords

• data clipping
• z-score normalization
• lorawan
• location accuracy
• rss