Research And Implementation Of Trajectory Data Query Processing On Cloud

In recent years, trajectory data goes everywhere, with the development and wide application of the device of mobile, and the location-based service. In the same time, a large amount of trajectory data has been generated, and become more and more important. The academic industry has been paying more and more attention to how to efficiently index, query, mine and analyze these large scale of trajectory data deeply, and it has become an important subject in its field. Because of the lack of expansibility, the traditional centralized trajectory data query process can’t handle this large scale of data. However, Hadoop has become a mainstream platform for big data due to its ability of efficient parallel processing. In this thesis, based on Hadoop, we study deeply the issue of querying of large scale of trajectory data. Our main works as follows:First of all, we propose a trajectory compression algorithm after studying and analyzing spatial-temporal characteristic of trajectory data. Based on the idea of spatial vector, this algorithm combines the method of data compression on plane with the characteristic of trajectory data, which can greatly reduce the cost of storage and indexing. Based on MV3R-tree, this thesis proposes an algorithm of optimizing of splitting of the leaf nodes of index tree, which can increases the similarity between entries in a leaf node, but decreases the similarity between leaf nodes. So the structure of index gets more collected.Then, we design two kinds of trajectory query algorithms based on Map-Reduce, mainly including timestamp query and range query. Besides, we propose an algorithm of path recommendation based on Dijkstra shortest path algorithm. This algorithm takes full advantage of trajectory information to search the recommending path which is reachable and quick. It is beneficial attempt to extend the analysis and application of trajectory further.At last, in order to test the performance and verify the validity of these algorithms proposed, we perform a lot of experiment. The result shows that, the trajectory data compression algorithm can decrease the cost of building and querying. Besides, the optimized index structure has better performance on parallel query processing, and it performs better than the centralized index. Also, the algorithm of path recommendation is good at small region query in term of responding speed and reachability.