Investigation of positioning accuracy based on received signal strength indicator (RSSI) values from radio waves

With the development and civilian deployment of GPS by the United States, many countries around the world have realized the strategic and commercial importance of having positioning systems. This has led to the emergence and increased use of various satellite-based position systems operating on a regional or global scale. However, the biggest disadvantage of these systems is the need for a clear line of sight between the satellite and the user on the ground. In other words, when using these systems, an obstacle (tree, wall, etc.) between the satellites and the receiver on the ground can greatly hinder the use of the system. For this reason, alternative systems and methods are being developed to meet the need for positioning in underground and confined spaces. One of these methods is positioning with RSSI. In this study, experiments have been carried out to determine the location in confined spaces using RSSI value and machine learning algorithms and to examine and improve the accuracy of the obtained position. In the study, Trilateration and Fingerprint methods were used as positioning methods. Due to the biggest disadvantage of the trilateration method, which is the effect of the placement geometry of the reference points on the calculated position accuracy, the reference RF transmitter stations used in the study were placed in two different geometries in the experimental field. For both geometries, RSSI and coordinate values were measured at points spread homogeneously over the experimental field and training data were collected and data sets were created for machine learning based regression models. In addition, RSSI values were measured with 10 epoch measurements and filtered with statistical methods. In this way, RSSI measurements that may be inaccurate due to environmental effects were tried to be eliminated. The models trained with this training data were hyper parameter optimized with an algorithm developed to increase the fit to the training data. In this way, it is aimed to obtain the highest performance from each model used in the study. The models with the highest accuracy were selected from the models whose parameters were adjusted and added to the prepared mobile application and real-time tests were performed for both geometries in the application area. For the distance and cooridnate values obtained as a result of the tests, methods to increase the position accuracy obtained by balancing with the EKK method were investigated. As a result, the method that gives the highest position accuracy values is selected and the parameters of the system for positioning with RSSI are proposed for the application area. It is thought that the developed method can be successfully applied in real application areas since the selected study area is closed, has limited visibility and physical obstacles similar to real world conditions.