Research On Major Technology Of Using WLAN And Fingerprint For Indoor Positioning

As the proliferation of wireless network and mobile devices continues, the requirement for Location Based Services (LBS) at the application level shows one of the most rapid and sizable growth trends. LBS is being developed so rapidly that it is being applied in many of social activities, and the LBS possesses great potential and an enormous market. As the requirement for location information increases, the development of localization technology and LBS are becoming linked together. The reliable and efficient indoor positioning technology is not only a prerequisite but also a critical factor for LBS.Due to the need for extra hardware, the overhead of the state of art of indoor position technology is pretty high. Furthermore, the hardware restriction exerts a significant effect on localization accuracy and coverage area, so this is a disadvantage for indoor applications of LBS. The Wireless Local Area Network (WLAN) and the Received Signal Strength (RSS) indoor positioning methods take advantage of the WLAN infrastructure. These kinds of methods satisfy the localization as a mobile terminal using some special software instead of another dedicated device. The WLAN based indoor positioning technology has some advantages due to its lower cost, and also due to the fact that it can almost satisfy all the accuracy demand of indoor positioning applications. However, with the widely deployed of Access Point (AP) as well as the changing of indoor intelligent terminal devices, the indoor radio propagation environment is increasingly complex. The RSS value shows high variability and large complexity due to the influence of the multipath effect, body absorption and signal interference. These affect the localization accuracy of WLAN fingerprint positioning based on RSS greatly, and presents challenges for WLAN fingerprint positioning technology.In this work, the RSS based WLAN fingerprint positioning technology is investigated and studied deeply. Aiming at practicability and application in LBS services, centred on the key problem of RSS reliability improving, the topic of how to improve the reliability and the availability of indoor positioning is discussed mainly in terms of theoretical aspects. Both hardware and software are utilized and many simulations and experiments have been performed. The major contribution of this dissertation as follows:· The distribution characteristics of indoor RSS are investigated. To describe the RSS distribution better, four classical indoor environments (common residence, office building, school building and marketplace) are adopted. The effects of people, receiver orientation and sample size to RSS signal are analyzed. A novel localization algorithm based on an Improved Double-peak Gaussian Distribution (IDGD) is proposed. Simulation and experiment results show that the localization accuracy of IDGD algorithm is quite good, while the sample size can be reduced greatly compared to traditional histogram and Gaussian-model based methods. Thus, the positioning overhead is comparatively low (saved70%samples). So, the IDGD algorithm addresses the problem of prompt indoor positioning in a small area comprehensively.· The classification problem in a large area is investigated. The statistical property of the RSS is changed even more in a large indoor environment. The computation complexity would increase if all the area must be studied. Furthermore, the optimal positioning model cannot be constructed easily, and the localization accuracy is poor. The computation complexity can be reduced and the localization accuracy can be improved by adopted the clustering algorithm to construct a sub-area model in a divided positioning sub-area. Aiming at improve the classification accuracy, the correlation of the signal is introduced, and an improved K-means clustering algorithm is proposed. Experiments indicate that the classification accuracy of the proposed algorithm is better (improved3.7%), the localization accuracy is high, the computation complexity is low and the algorithm performs well in terminal energy saving.· The access point (AP) selection problem is studied. The information quantity of different AP is distinct, especially in the circumstance that public APs are deployed densely. Many RSS values are influenced by various sources of noise and contain lots of redundancy. Thus, these RSSs do not contribute to improving localization accuracy but reduce it instead. So not all of the RSSs are beneficial for localization. Determining the localizability is necessary to select the optimal set of APs for localization. The state of art of AP selection has not considered the recall ratio and precision ratio; an information gain using an AP selection algorithm is presenedt. Experiments demonstrate that the algorithm can get rid of the redundant APs to improve localization performance.· The RSS positioning feature extraction issue is explored. Extracting the RSS feature is beneficial to getting rid of the RSS signal redundancy and to improve the RSS signal reliability. In view of the existing methods to take account of extracting the RSS linear feature only, a Kernel Discriminant using Linear Discriminant Analysis (KD-LDA) algorithm is formulated. The proposed algorithm takes full advantage of the non-linear feature of the RSS. Experiments prove that, since the clustering algorithm and the AP selection method are combined in the SVM based regression positioning model, the probability of positioning with low error is increased along with a reduction in the error range leading to enhanced localization performance.