Radio Map Augmentation Using DBSCAN and KNN Regression for Improved Indoor Positioning

Abstract

The proliferation of smartphones has driven an increased demand for indoor location-based services. Consequently, location fingerprinting with Received Signal Strength Indicators (RSSI) has become a popular method for indoor positioning, as it accurately determines a target location by effectively mitigating the multipath effect commonly encountered in indoor environments. However, the fingerprinting technique faces challenges in constructing a radio map, which is exceedingly time-consuming and labor-intensive, limiting its real-world application. To address this issue, synthetic RSSI data can be generated using small datasets collected from sparse reference points (RPs). This paper proposes a method for generating synthetic data using the KNN Regression approach. To improve the accuracy of synthetic data synthesis, we employed DBSCAN to partition the entire region into clusters. We evaluated our proposed method using a radio map collected from 18 Wi-Fi routers in a two-story university building, reducing the radio map by uniformly removing data from several RPs by 25%, 50%, and 75%. The method was then applied to enhance the incomplete datasets. The results indicate that the proposed method successfully reduced the average positioning error to 0.202 meters for the 75% reduced radio map, 0.048 meters for the 50% reduced radio map, and 0.116 meters for the 25% reduced radio map.