Enhanced ANDSF WiFi Discovery Mechanism Using Machine Learning for Mobile Data Offloading

Mobile data offloading using unlicensed WiFi spectrum represents one of the innovative approaches that can be adopted to cope with the huge amount of mobile traffic demand. However, the main challenge in such approach lies in the ability of the system to discover the WiFi coverage for users who may need to offload their data. For this, the access network discovery and selection function (ANDSF) has been introduced by 3GPP to provide the mobile user equipment (UE) with the discovery information and the list of available WiFi hotspots. However, the current ANDSF discovery mechanism depends on the geographical location of the UE, which need to be accurately determined and continuously transmitted to the ANDSF server. In this paper, an enhanced ANDSF WiFi discovery technique is proposed and validated. The proposed technique is designed to use the reference signal received power (RSRP) information measured at UEs from their surrounding cellular base stations to build fingerprints for various WiFi hotspot areas. Based on the constructed fingerprints, the UE's WiFi coverage state can be identified using the proposed decision tree (DT) machine learning approach. The conducted tests and generated results validate the operation and performance of the proposed machine learning-based discovery technique with accuracy reaches up to 95%.