Indoor localization based on principal components and decision trees in IEEE 802.15.7 visible light communication networks

Abstract

Indoor positioning estimation has become an attractive research topic due to the growing interest in location-aware services. Research works have been proposed on solving this problem by using wireless networks. Nevertheless, there is still much room for improvement in the quality of the proposed classification or regression models, i.e., in terms of accuracy or root mean squared error (RMSE). In the last years, the emergence of Visible Light Communication brings a brand new approach to high quality indoor positioning. Among its advantages, this new technology is immune to electromagnetic interference, and also, the variance of the received optical power is smaller than other RF based technologies. In this paper, we propose a fingerprinting indoor location estimation methodology based on principal components analysis (PCA) and decision trees as classification learner. The proposed localization methodology is based on the received signal strength from a grid of emitters multiple. PCA is used to transform all of that features into principal components, consequently reducing the data dimensionality, improving the interpretability of the resulting tree models and the overall computational performance of the proposed system. Along with the proposed method, we also share experimental results derived from the received signal strength values obtained from an IEEE 802.15.7 simulator developed by our research group. Results show that the system accuracy is slightly improved by range 1%-10% and the computation time by range 40%-50%, as compared to the system in which PCA is not carried out. The best tested model (classifier) yielded a 95.6% accuracy, with an average error distance of 2.4 centimeters.