Enhanced RSSI Based Indoor Positioning System by a Fusion of Multiple Machine Learning Algorithms

Tracking an object in the indoor environment has attracted researchers due to the increased randomness and interference compared to the outdoor environment. These factors reduce the precision of popular localization algorithms in an indoor environment. This paper proposes a fusion of three machine-learning methods to address this issue. This method combines feedforward neural network (FFNN), eXtreme gradient boost (XGBoost) and support vector machine (SVM) to estimate location over a set of WiFi location fingerprints. The experimental results show that the method significantly improves the precision of the output position in both case studies, with high test accuracy. Nevertheless, the paper's limitation lies in the assumption of a consistent environment or with objects moving under 1 m/s.