Investigation of Beacon-based location determination methods indoor in terms of accuracy and usability

Many technologies are being developed and used for positioning in indoor. One of these technologies is Beacon technology with BLE (Bluetooth Low Energy) infrastructure. Within the scope of this thesis, the performance of Beacon technology for positioning purposes in indoor where GNSS (Global Navigation Satellite Systems) technology is insufficient was investigated. In studies on positioning in indoor spaces, the selection of the positioning method, its implementation and positioning performance along with the equipment to be used are very important. In the scope of this thesis, proximity, trilateration and fingerprinting methods, which can be used for indoor positioning are closely examined together with Beacon technology. Within the scope of the study, a simple Beacon-based navigation application in a single-storey indoor and a Beacon-based navigation application in multi-storey indoor which is an improved version of this application, were designed and tested. As a result of the tests, 95% success was achieved in multi-storey indoor spaces, considering the target reaching situation determined for the Beacon-based navigation application. For the proximity method which is one of the positioning methods used in indoor, distance values obtained automatically from beacons and distance values calculated and obtained according to the ambient conditions were compared within the scope of this thesis. Polynomial interpolation and ANN (Artificial Neural Networks) were used for the distance value calculated according to the ambient conditions and Kalman filter was used to eliminate the noise in the Beacon signals. As a result of the tests, the average distance values between the Beacon and the user were calculated with an average accuracy of 0.95 m using 6th degree polynomial interpolation. When the data obtained as a result of this process was evaluated with the Kalman filter, the average distance values between the Beacon and the user were calculated with an accuracy of 0.82 m. The same distance values were calculated with an average accuracy of 1.73 m with the distance value obtained automatically from the Beacons. For the trilateration method, one of the positioning methods used in indoor, tests were carried out in two different indoor. In the first indoor tests, the location was calculated according to six different positioning schemes, taking into account the Beacon sequences. In these calculations, the classical trilateration method (using the three strongest Beacon signals) and the least squares trilateration method (when the number of Beacon signals is above three) were used. In the tests, the position was calculated by trilateration method using the distance value obtained automatically from the Beacons and the distance value calculated according to the ambient conditions. As a result of the study, the position was calculated with a maximum error of 1.01 m for the test points within the square for 3x3 m and 4x4 m square schemes. In the tests performed for the trilateration method in the second indoor and in a 4x6 m area, the location was calculated with eight different combinations, including the classical trilateration method, the trilateration method with LSQ (Least Squares), the trilateration method using the Kalman filter, and the method of obtaining the distance between the Beacon and the user. The trilateration method, in which A and n values (parameters used for the distance value calculated according to ambient conditions) for the distance between the Beacon and the user are calculated and used together with the Kalman filter, gave the best location accuracy with an average location error of 1.98 m. For the fingerprint method, which is one of the positioning methods used indoors, the position was calculated by considering many situations. These situations consist of factors such as the number of Beacons to be used for the fingerprint method according to the size of the indoor, data collection direction, data collection intervals, data collection heights, number of data collection points, classification methods to be used, the use of Kalman filter to eliminate noise in Beacon signals. NN (Nearest Neighbors), KNN (K-Nearest Neighbors), WKNN (Weighted K-Nearest Neighbors) and ANN were used as classification methods in the studies conducted for the fingerprint method. In the tests conducted for the fingerprint method, the ANN method gave the best performance with a location error of 1.77 m using the Kalman filter, based on the average location accuracies.