Aziz, T.; Camana, M.R.; Garcia, C.E.; Hwang, T.; Koo, I. REM-Based Indoor Localization with an Extra-Trees Regressor. Electronics2023, 12, 4350.
Aziz, T.; Camana, M.R.; Garcia, C.E.; Hwang, T.; Koo, I. REM-Based Indoor Localization with an Extra-Trees Regressor. Electronics 2023, 12, 4350.
Aziz, T.; Camana, M.R.; Garcia, C.E.; Hwang, T.; Koo, I. REM-Based Indoor Localization with an Extra-Trees Regressor. Electronics2023, 12, 4350.
Aziz, T.; Camana, M.R.; Garcia, C.E.; Hwang, T.; Koo, I. REM-Based Indoor Localization with an Extra-Trees Regressor. Electronics 2023, 12, 4350.
Abstract
As an established and widely available infrastructure, wireless local area networks (WLANs) have emerged as a viable option for indoor localization of both mobile and stationary users. In planning a mobile communications network and radio system design, the critical role of coverage prediction becomes evident, empowering network operators to optimize cellular networks and elevate the overall customer experience. Moreover, WLANs present several challenges that must be fulfilled when it comes to localization based on Wi-Fi signals to get a proper coverage prediction map. This paper presents a study based on application of the extra trees regression (ETR) for indoor localization by using coverage prediction maps. The aim of the proposed method is to accurately estimate a user’s position within a radio environment map (REM) area using collected received signal strength indicator (RSSI) values. The proposed scheme investigates several machine learning (ML) regression algorithms for localization, where the training dataset is obtained from the REM by using the nearest neighbors method. Parameter tuning is conducted to optimize the performance of the ETR scheme by using 10-fold cross-validation. In the numerical results, we first demonstrate the effectiveness of utilizing ML regression techniques for generating coverage maps, which enables accurate estimation of the Wi-Fi signal strength in indoor environments. Then, we showcase the superior performance of the proposed ETR-based method compared to several other ML schemes for indoor localization using the REM. ML algorithms, including decision tree regression and the ETR, are compared to evaluate the system model. Based on error metrics, the proposed ETR-based approach exhibits the best performance among the evaluated techniques. The combination of coverage map generation and localization using regression techniques offers a powerful approach for analyzing the radio frequency (RF) environment in indoor spaces.
Keywords
machine learning; indoor localization; radio environment map; extra trees regressor; cross-validation
Subject
Engineering, Electrical and Electronic Engineering
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
