چكيده به لاتين
Location data is one of the most important context data in the Internet of Things. Since the global positioning system doesn’t work properly inside buildings, indoor positioning systems are developed to provide location-based services in these environments. One of the common ways to implement these systems is to use radio waves, like Wi-Fi. However, the complexity of the propagation of waves indoors and, as a result, the fluctuations of received signals, made the development of error reduction methods a necessary action.
In this thesis, methods have been developed to increase the accuracy and reduce the computational complexity of indoor positioning systems within multi-floor buildings. These methods, which are presented under the fingerprint grouping scheme, limit the search space of fingerprint reference points or change their factors. The plan is divided into three sub-schemes.
In the first sub-scheme, we cluster the reference points based on the vicinity of the access points (i.e., the access point with a signal stronger than a determined threshold). In the online phase, by taking the signal strength vector, it is possible to quickly identify the cluster(s) and perform the positioning in the same cluster(s). In the second sub-scheme, we reduce the search space based on the last calculated position. In the third sub-scheme, we also use the environment map as a graph to use walking constraints in the building, especially in changing floors. The presented methods are evaluated using field tests. The first sub-scheme reduced the average error by about 5%. The second sub-scheme also reduced the positioning error in one floor about 1 meter, and the third sub-scheme reduced floor recognition and resulted in reducing the mean error by 2.4 m. Also in the evaluation of execution time, reductions from 1.5 to 10 times have been reported.
