TraSP: A General Framework for Online Trajectory Similarity Processing

Trajectory similarity is one of the most fundamental operations in spatial-temporal data analysis. Although many recent works focus on improving the efficiency on single machine, their solutions are not directly applicable to DSPEs (Distributed Stream Processing Engine) in an online manner. On one hand, the similarity processing on DSPEs is always susceptible to data skew and completeness issue. On the other hand, their methods only support a single trajectory similarity measure which could not serve for adaptive adjustment strategies in different scenario. In this paper, we propose a new general framework for online Trajectory Similarity Processing, named TraSP. Specifically, our proposal includes a matrix-based data dispatcher to provide balance and completeness guarantee for stream join, an atomic table generator to accommodate different similarity criteria and a lightweight filter to shed irrelevant workloads. Empirical studies on real world trajectory data sets validate the usefulness of our proposals and the comparison experiment shows the high performance of our framework.