PROJECT TITLE :

Cloud-based Framework for Spatio-Temporal Trajectory Data Segmentation and Query

ABSTRACT:

The process of separating a sequential trajectory into individual segments is referred to as trajectory segmentation. These segments are the fundamental constituents of a variety of applications. As a result, a system framework is necessary in order to support the indexing, storage, and querying of trajectory segment information. A distributed solution is proposed whenever the size of the segments exceeds the capacity of a single processing node to perform the necessary computations. In this piece, we develop a framework for distributed trajectory segmentation that incorporates a greedy-split segmentation method. This framework is made up of distributed processing that takes place in memory and a cluster that stores graphs respectively. We devised a distributed spatial R-tree index of trajectory segments so that fast queries could be performed on the trajectory. With the help of the indexes, we are able to construct scalable query operations from in-memory processing as well as access to graph storage. On the basis of this framework, we define two metrics in order to measure the similarity of the trajectories and the likelihood of collisions. These two metrics are then used in conjunction with one another to determine which moving groups of trajectories exist. On the system, we perform a quantitative analysis to determine the effects of data partitioning, parallelism, and data size. We identify the factors that are causing the bottleneck at the data partition stage and validate two techniques that can be used to mitigate data skew. The evaluation demonstrates that both our distributed segmentation method and the system framework are scalable in response to an increase in the amount of work being performed in parallel as well as the size of the cluster.