Driving behaviour analysis of spatiotemporal trajectories by using appropriate similarity measures.

Kurzfassung

This work focuses on similarity measures that can be applied on different driving behaviors. Also methods are shown how similarity measures can be implemented. An overview about different kind of distance operators is given and later the distance operators were categorized. This work is based on data from the UR:BAN project founded in the year 2012. Two use cases were defined which can be uses to do further research using similarity measures. An introduction to spatial and spatiotemporal trajectories is given. The characteristics of these kind of trajectories are defined and compared to time series. Seven shape based distance operators are introduced as well as four edit based ones. Examples for edit based distance operators are EDR and ERP, while the shape based include the Locality in between Polylines distance operator which also is used for implementation. Different algorithms are used to add spatiotemporal functionality to selected distance operators. The Bentley-Ottmann algorithm is used for finding polyline intersections in spatiotemporal trajectories and the Douglas-Peucker algorithm is used for thinning created polygons. In the end two new R packages are developed for further use. This includes either one spatial and one spatiotemporal distance operator. The packages are developed in C++ . The basic algorithms for selecting data and performing calculations were written in R, a statistical programming language. As data source a simulator test run database is used.