Indoor Positioning System has been one of the most attractive research after Bluetooth Low Energy (BLE) was introduced. This technology mainly used because of the reduction of material and energy cost over time that has huge impact compared to other technology which are more costly. Most Recent Research has solved the problem of inconsistency on Received Signal Strength (RSS) using fingerprint method. But the RSS that received may contain some noise. This paper mainly proposed a method to estimate or tracking the real position of dynamic user. With RSS value as input to be processed and the result of it will be a location (x, y) value, then repeat the process to create an estimate coordinate map of route taken. Our proposed method is based on fingerprinting with weighted sum of five nearest reference point to estimate the position of dynamic user then using Extended Kalman Filter as a tracking algorithm. In this paper we try new ways to collect the data of RSS by dynamically collecting the data in many routes to see whether the proposed algorithm could estimate the position better. We achieve an average mean of error around 302,42 cm using Weighted Sum + Extended Kalman Filter tested on dynamic data.

Indoor Positioning System has been one of the most attractive research after Bluetooth Low Energy (BLE) was introduced. This technology mainly used because of the reduction of material and energy cost over time that has huge impact compared to other technologies, which are more costly. Most recent research resolve around improving the accuracy of calculated position of the user by implementing different method to enable an indoor positioning system, and to remove any noises in the dataset. This paper objective is to compare some of the available methods that are used to enable Indoor Positioning System such as Fingerprinting, Multilateration, Trilateration, and Heron Bilateration. Since the performance of Fingerprinting is better compared to other methods, we combine Fingerprinting’s offline phase with the other methods to create a hybrid method and compare the accuracy of predicted user’s position. The experimental results show that the Fingerprinting and WKNN method outperform all other methods by resulting on 271.76 cm mean of error.