Neural Fuzzy Based Indoor Localization by Extending Kalman Filtering with Propagation Channel Modeling

In this study, a low complexity WLAN indoor positioning system based on the curve fitting, the interpolation, and an adaptive neural fuzzy system is addressed. The proposed modeling techniques are used to reduce the false distance estimation of a mobile device to the access points induced by the unpredictable interference, reflection and multipath effects. It is clear that the experimental results show that a empirical propagation model can be easily provided by the curve fitting based approach but involves a limited flexibility and adaptation; the interpolation based approach has a simple computation for the distance estimation and a high flexibility due to the possible arragement of the reference nodes but requires horrible trails and errors for optimization. Therefore a hybrid learning algorithm under a combination of gradient descent and the least-squares method based neural fuzzy inference system is proposed to solve the problems mentioned above. Due to its nonlinear ability, fast learning capacity and adaptation capability, the accurate distance estimates of a mobile device to the access points can be obtained. The distances are firstly converted by the addressed ANFIS propagation model during the the online stage in real-time with the currently observed RSSI value from the tested WLAN area. The available distances from the positions localized in the test bed are recursively estimated by an extended Kalman filter while it could blend the information optimally minimizing the variance of the estimation error. Consequently, the presented approach is verified to provide the accurate localization in the real indoor environment of 20.6m×37.4m with 4 access points only. The addressed positioning algorithm is straightforward and can be implemented by simple mobile devices, such as PDAs or mobile phones, to provide the location-aware services without any hardware cost since WLAN signals are popular sensor sources for indoor environments. There are many potential applications for location positioning and network accessing, such as library or museum tour-guide, free mobility and nursing at home, patient transporting in the hospital and easy going capability in the shopping mall.