Vol.8 (2018) No. 4-2 ISSN: 2088-5334 Wi-Fi Indoor Positioning Fingerprint Health Analysis for a Large Scale Deployment KS Yeo *1 , A Ting #2 , SC Ng #3 , D Chieng #4 , N Anas #5 * Department of Electrical and Electronics Engineering, Faculty of Engineering and Technology, Tunku Abdul Rahman University College, Jalan Genting Kelang, Setapak, 53300 Kuala Lumpur Email: yeoks@tarc.edu.my # Wireless Innovation, MIMOS Berhad, Technology Park Malaysia, Bukit Jalil, 57000 Kuala Lumpur Email: #2 kee.ting@mimos.my; #3 sc.ng@mimos.my; #4 ht.chieng@mimos.my; #5 nuzli.anas@mimos.my Abstract - Indoor positioning systems (IPS) have witnessed continuous improvements over the years. However, large-scale commercial deployments remain elusive due to various factors such as high deployment cost and lack market drivers. Among the state of the art indoor positioning approaches, the Wi-Fi fingerprinting technique, in particular, is gaining much attention due to its ease of deployment. This is largely due to widespread deployment of WiFi infrastructure and its availability in all existing mobile devices. Although WiFi fingerprinting approach is relatively low cost and fast to deploy, the accuracy of the system tends to deteriorate over time due to WiFi access points (APs) being removed and shifted. In this paper, we carried out a study on such deterioration, which we refer to as fingerprint health analysis on a 2 million square feet shopping mall in South of Kuala Lumpur, Malaysia. We focus our study on APs removal using the actual data collected from the premise. The study reveals the following findings: 1) based on per location pin analysis, ~50% of APs belong to the mall operator which is a preferred group of APs for fingerprinting. For some location, however, the number of operator-managed APs are too few for fingerprinting positioning approach. 2) To maintain mean error distance of ~5 meters, up to 80% of the APs can be removed using the selected positioning algorithms at some locations. At some other locations, however, the accuracy will exceed 5m upon >20% of APs being removed. 3) On average, around 40% - 60% of the APs can be removed randomly in order to maintain the accuracy of ~5m. Keywords – indoor location positioning; fingerprint; Wi-Fi I. INTRODUCTION The transportation industry has witnessed an unprecedented disruption brought by location-based services such as Grab and Uber. Despite the success, the usage is only limited to outdoor as GPS signal is not able to penetrate through building structures. Realizing the gap, efforts have intensified in the development of an indoor positioning system in recent years. Its social and commercial values are estimated to worth around USD10 billion by 2020 [1]. Among the crucial areas that can benefit from indoor positioning systems include public safety (e.g., E911 emergency call, in building rescue), retail (e.g., store search in a shopping complex, mobile advertising, asset tracking), special care group (e.g., children and special disabled needs), etc. [2]. Various techniques have been proposed to build an indoor positioning system such as those based on Wi-Fi signals [3]– [5], Bluetooth signals [6], [7], FM radio signals [8], [9], RFID signals [10]–[13], sound waves [14], [15], light signals [16], [17] and magnetic field [18], [19]. Among these techniques, the Wi-Fi fingerprinting approach is gaining much attention due to two main reasons. Firstly Wi-Fi access points (APs) are widely available across most commercial and residential premises. This enables indoor location positioning system to be deployed rapidly with minimal cost. Secondly, Wi-Fi module is available in almost every consumer device such as a smartphone, tablet, laptop, and wearable. This serves as a great commercial advantage as there is no need to provide extra hardware to the mass users [20]. The deployment of a Wi-Fi fingerprinting indoor positioning system involves two phases: an offline phase (fingerprint calibration) and online phase (real-time positioning) [21]. During the offline phase, a site survey is conducted to record all APs within the scanning range and their associated received signal strength indicator (RSSI) at every reference point (location pin) of interest. With these records, the fingerprint database can then be constructed. During the online phase, a mobile device continuously scans 1411