Positioning accuracy of smartphones in crowdsourcing context

Smartphones have become an important aspect of our daily lives and are equipped with a wide range of features. One of the features is the ability to collect positioning data. The devices are mass-produced and widely used, which is why many organizations are interested in their positioning applications. Many of the studies have focused on examining a few smartphones. The main objective of this thesis was to study the positioning accuracies of smartphones using real-time and post-processing positioning techniques. In this case, the processes have not been optimized to produce the best positioning results. The interest lay more in determining positioning accuracy in crowdsourcing context. The field measurements were conducted all over Finland, where the environment varied from rural to urban areas. Real-time measurements were taken during the field sessions. Post-processing was done using two different software, which were RTKLIB and SSRPOST. Python scripts were developed to mass-process and analyse the large dataset. Positioning accuracy of individual measurements differed quite significanty depending on the used positioning technique. Real-time results were roughly at 10 metres, which is considered inaccurate. Post-processing produced more accurate results, where SSRPOST had the best accuracy at 3,92 metres. To improve the accuracy, centre points of smartphone measurements were determined and used in positioning calculations. Real-time accuracies improved roughly to 6 metres. From post-processing, SSRPOST produced the most accurate results at 1,46 metres. In addition, positioning accuracy change when the number of measurements increases was determined. The biggest improvement in positioning accuracy happened between measurements two and three. The positioning accuracies stabilised when around 10 measurements were used. Despite the positioning accuracies being less accurate than what was found in the literature, the developed processes have their applications. Using post-processing and centre points of measurements improved the positioning accuracies significantly. This process works best for applications that can collect thousands of measurements and an accuracy of couple of metres is enough