Scalable Indoor Wi-Fi Positioning Algorithm Using Cluster-Based Local Estimators

Most existing WiFi-fingerprinting indoor positioning methods only utilize a single machine-learning model to estimate the device coordinates using the radio signal strength database. However, these approaches may encounter scalability problems in large radio maps where the radio patterns vary significantly amongst different areas. To address this issue, this paper introduces a Clustered-Based Local Estimator (CBLE) algorithm that combines results from local and global estimators to better formulate radio signal strengths with the corresponding position coordinates. The proposed algorithm is then evaluated with some popular open-source datasets (UJIIndoorLoc and UTSIndoorLoc). The experimental results verify that our proposed approach effectively reduces the positioning error by ≈ 24% and 10%, respectively, for the UJIIndoorLoc and UTSIndoorLoc datasets in comparison to state-of-the-art methods.