Research On Wi-Fi-based Indoor Positioning Technology

Due to the rapid development of mobile communicate technology and the popularization of intelligent mobile phone, the indoor positioning systems has attracted more and more attention in Location-based Service regimes. However, the mature GPS (Globe Positioning System) could not be used for indoor environment. In recent years, according to the unique indoor environment, the Wi-Fi-based indoor positioning system has been extensively studied by the researchers, because they can be applied to wireless networks without needing any extra hardware facilities. Considering the existing indoor positioning systems, owing to the lower cost and higher positioning precision, the fingerprint-based indoor positioning system shows a clear advantage.In this dissertation, we have made a thorough research on fingerprint-based indoor positioning system. Due to the fluctuant characteristics of received signal strength and the accuracy of previous systems heavily relies on environmental conditions, it makes a huge challenge for fingerprint-based indoor positioning system. In view of these factors, the main work and innovations are as follows:At first, this dissertation introduces several widely used algorithms for the fingerprint-based indoor positioning system, such as KNN algorithm, Bayesian decision algorithm, PPMCC algorithm, neural network algorithm and SVM regression algorithm. For the real-time requirement of positioning technology, we have researched several clustering methods for fingerprint database which can effectively reduce calculation in real-time positioning.Secondly, this dissertation proposes a novel Multi-classifier-Based Multi-Agent model for Wi-Fi positioning system. By the output layer classifier fusion, it can output optimal position of each classifier. According to the experiments of the multi-classifier performed in two different environments, the combination for multiple numbers of classifiers could significantly mitigate the environment-dependent characteristics of the classifiers. The average error distances of the multi-classifier was found to superior to that of the other single classifier in two test environments.At last, we present a test environment for Wi-Fi based indoor localization, and analyze the static positioning performance for KNN algorithm, Gaussian distribution algorithm and PPMCC algorithm. Through using different training and testing equipment, the experimental results show that the PPMCC algorithm can effectively solve device heterogeneity problem. At the same time, we utilize the singular value calibration and Kalman filtering to smooth the jitter feature for the dynamic positioning.