An Enhanced RSS Fingerprinting-based Wireless Indoor Positioning using Random Forest Classifier

Fingerprinting-based wireless indoor positioning which utilizes Received Signal Strength (RSS) data achieves increasing attention for modern location-based services due to its benefits of cost effectiveness and ease of deployment. Recent research works in RSS fingerprinting based wireless indoor positioning employed Random Forest (RF) classifier, which is one of the most popular machine learning techniques, to achieve outstanding localization performance. However, there are still some shortcomings in the existing works due to high RSS variability and multipath fading effects in indoor environments. To further improve the performance of existing works, an efficient RSS fingerprinting wireless indoor positioning using Random Forest classifier is proposed, in which RSS data from two different frequency bands: 2.4 GHz and 5GHz are deployed in building the fingerprinting database. In addition, RSS readings are taken with four different antenna orientations at each reference point to improve positioning accuracy and mitigate effect of radio irregularity. Experiments are conducted in a real- world test-bed and results indicate that the proposed Random Forest (RF) based wireless indoor positioning system achieves 1.69-meter precision with improved positioning accuracy than existing RF based approaches.