Efficient pattern query processing over trajectory data

An increasing volume of data as sequences of locations, called 'trajectories', are being generated every day by moving objects equipped with GPS devices. The availability of such data has driven several studies on trajectory query processing in the last two decades. While most of the existing studies on trajectory queries consider the trajectories independently, the queries that consider the interrelations among trajectories to return multiple trajectories with similar movement as a pattern, convey useful findings for many real-life applications. Towards this direction, this thesis focuses on the efficient processing of the pattern queries on spatio-temporal and semantic trajectory data. Specifically, we study the following three problems: (i) multi-range query over spatio-temporal trajectory data; (ii) recurrent convoy query over streaming trajectory data; and (iii) top-k semantic trajectory pattern query.

The multi-range query is motivated by the observation that trajectories generated by moving objects can be used to explore the movement dynamics among different areas over the search space during a specific time period. Such information is beneficial for many real-life applications, for example, urban planning and intelligent infrastructure management. In this thesis, we study the problem of efficiently finding the trajectories that pass through all of the given spatio-temporal query ranges.

One of the interesting findings over trajectory databases is the exploration of co-moving patterns that are generated by objects travelling together for a certain period of time, called convoys. Some applications require to find the repeated occurrences of such patterns. An example application is to find the recurrent occurrences of traffic congestions for planning purposes, and also to distinguish unusual traffic congestions from the regular recurrent ones. Moreover, as the traffic data continuously changes due to the moving vehicles, the reoccurrences of such patterns need to be continuously checked for the new incoming data. In this thesis, we address the problem of finding recurrent convoys over a sliding window of spatio-temporal trajectory data. We focus on efficiently finding the convoys in the new incoming data that are repeated occurrences of convoys in historic data.

Trajectories that pass through locations with semantic information allow users to explore the historical travel movement patterns of the other travellers by specifying the keywords of interest. These patterns are potentially useful for applications such as travel recommendation and route planning. As the number of such patterns can be large, ranking and finding the most informative ones are beneficial for most applications. In this thesis, we study the problem of finding top-k semantic trajectory patterns where each resulting pattern is formed by a set of similar trajectories that pass through locations associated with the query keywords.