Parallel trajectory search based on distributed index

With the massive location-based information contained in public available GPS enabled mobile devices, there are problems of storage and queries processing. Trajectory search processing has received increasing attention in recent years. To organize the huge information, many studies have used indexing large trajectory data sets. However, the majority of existing studies have focused on a centralized system, while Spatial Hadoop is proposed to handle the spatial data by MapReduce method in a distributed system. However, this method is not suitable for reusing data without storing it on a disk. Therefore, we have developed a new R-tree index in the distributed system called the Distributed Trajectory R-Tree (DTR-Tree) with Apache Spark based on the MapReduce model. Furthermore, we have investigated a novel problem of a distributed trajectory query search with activities. Where a query q is given with a set of activities involved in each point of the trajectory and a threshold of distance, then the trajectory, which includes all the activities required with a minimal distance, is returned as a result. To optimize the trajectory processing search, we have used a novel strategy on separated smaller R-trees to obtain the sub-trajectories matching by q based on the spatial distance. In this paper, a balanced distributed index DTR-Tree is proposed to ensure the scalability and fault tolerance. Experimental results show the high efficiency of the proposed algorithms.