Noggrannheten hos ett positioneringssystem baserat på Bluetooth 5 : En jämförande studie av fem olika filtreringsalgoritmers påverkan på positioneringsnoggrannheten

Sammanfattning: Purpose – The purpose of this thesis was to identify and evaluate different filtering algorithms for raw Received Signal Strength Indication (RSSI) values in an indoor positioning system. The purpose was also to examine how the positioning accuracy is affected depending on which filter that was applied to an indoor positioning system based on Bluetooth Low Energy (BLE) 5. Method – To identify and evaluate the filtering algorithms a literature study was performed, where databases such as Primo, Google Scholar, Digitala Vetenskapliga Arkivet among others were used to collect data. To examine how the positioning accuracy is affected by the filtering algorithms, an experimental study was performed. A one-dimensional indoor positioning system was designed were a Blue Gecko BGM111 Bluetooth Low Energy Model with BLE 5 was programed to read the RSSI values from a connection with an iPhone 8. The RSSI values were sent via a UART to an external computer which stored these as a text file. All the filtering algorithms processed the same text file to reassure comparable conditions. The results from the different filtering algorithms were examined in two steps. The first step was to decide if the filtering of the distance values or the filtering of the RSSI values resulted in the highest positioning accuracy. The second step was to compare the results to decide which of the filtering algorithms that gave the highest positioning accuracy. The examination resulted in an observed hypothesis created by the authors, that was validated by conducting a single tailed hypothesis test with a significance level of 0,01. By conducting the tests in three different environments, an assurance was created that the same phenomenon was in place in all environments. The experimental study took place in the following environments: harsh office environment, calm office environment and open environment. Findings – The results from the literature study showed that the Kalman filter, Feedback filter, Gauss filter, Moving Median filter, Moving Mean filter and Particle filter are the most frequently mentioned filtering algorithms in the positioning context. The results from the experimental study showed that the Kalman filter was the filtering algorithm with the highest positioning accuracy. The results also showed that filtering on the RSSI values before they were transformed into distance was an improvement compared to filtering on the distance, after the transformation of the RSSI values. Implications – The results from this thesis will be beneficial for ROL Ergo AB in their upcoming implementation of an indoor positioning system. After a conducted examination of existing work within the field, the authors found a lack of studies that compare positioning accuracy on more than 3 filtering algorithms and with a usage of the latest version of BLE, namely version 5. This shows that the results from this thesis could be of great use for developers of positioning systems based on BLE. Limitations – The time spent realizing this thesis did not allow for a full and exhaustive evaluation of the great number of filtering algorithms that are available. Therefore, only a few filtering algorithms were selected to be examined and tested, which was done by what the authors considered to be a frequent mentioned filtering algorithm in other positioning systems and that was implementable within the timeframe of this thesis. This thesis experimental study only used BLE 5 as communication mean and therefore does not contain a comparison of positioning accuracy between BLE 5 and earlier versions of BLE. Keywords – RSSI, BLE 5, filtering algorithm, positioning accuracy.