Recently, the static and dynamic geo-location of a device or a person has become one of the most important aspects of communication systems because of its multiple applications. In general, knowing the position of a mobile terminal (MT) in outdoor or indoor environments is of major importance for applications providing services based on the location. The development of localization systems has been mainly driven by the avail- ability of the affordable cost of indoor wireless local area network (WLAN) infrastructure. There exist different techniques to localize MTs with the different mainly depending on the type of the metrics extracted from the radio frequency signals communicated between base stations (BSs) and MTs. Ideal measurements are taken in environments which are free of obstacles and in direct ray tracings between BS and MT. This is not the case in indoor environment because the daily use of permanent obstacles in the work space scatters the ray tracings. Measurements taken in Non Line Of Sight (NLOS) are unpredictable and different from those taken in LOS. In order to reduce measurement errors, one can apply different techniques such as mitigation, approximation, prior correction, or filtering. Tracking systems (TSs) have many concrete applications in the space of individual navigation, social net- working, asset management, traffic management, mobile resource management, etc. Different techniques are applied to build TSs in indoor environments, where the signal is noisy, weak or even non-existent. While the Global Positioning System (GPS) devices work well outside buildings and in urban canyons, tracking an indoor user in a real-world environment is much more problematic. The prediction problem remains an essential obstacle to construct reliable indoor TSs. Then lacks of reliable wireless signals represent the main issue for indoor geo-location systems. This obviously calls for some sort of predictions and corrections to overcome signal reliability, which unavoidably open the door for a multitude of challenges. Varieties of approaches were proposed in the literature. The most used are the ones based on prediction filters, such as Linear Filter (LF), Kalman Filter (KF) and its derivatives, and Particle Filters (PF). Prediction filters are often used in estimation problems and applying Digital Fractional Differentiation can limit the impact of performance degradations. This work presents a novel approach for the WLAN indoor geo-location by using coordinates clustering. This approach allows overcoming the limitations of NLOS methods without applying any of mitigation, approximation, prior correction, or filtering approaches. Then a comparison study of deterministic and learning techniques for indoor geo-location is presented. Finally, it presents a novel soft approach for indoor tracking system by applying digital fractional integration (DFI) to classical prediction filters
