Accurate WiFi-based Indoor Localization by Using Fuzzy Classifier and MLPs Ensemble in Complex Environment

Abstract With the rapid increase of mobile devices, there are a lot of location-based applications. Therefore, the localization of indoor environments is an increasingly important problem. WiFi-based fingerprint method which is cost-effective without investing additional infrastructure has drawn significant attention over the past decade. However, due to the interference of moving objects and so-called co-channel interference, incurring the high variability of WiFi signals over time for the same location and make it hard to obtain the satisfied accuracy and mean error. To remedy those problems, in this paper, we propose an ensemble model consisting of fuzzy classifier and multi-Multi-layer perceptron (MLPs) for indoor parking localization. The clustering algorithm is applied to get the similar areas and form the local model, and then the ensemble learning is trained in offline stage. In online stage, the ensemble learning is utilized to get real position of unlabeled input. The experiment has been conducted at indoor parking of Riyueguang mall in Chongqing, China. This proposed approach yields higher accuracy and lower mean error, and makes it possible to apply WiFi-based localization in real indoor parking.