Spatio-temporal trajectory simplification for inferring travel paths

Mining GPS trajectories of moving vehicles has led to many research directions, such as traffic modeling and driving predication. An important challenge is how to map GPS traces to a road network accurately under noisy conditions. However, to the best of our knowledge, there is no existing work that first simplifies a trajectory to improve map matching. In this paper we propose three trajectory simplification algorithms that can deal with both offline and online trajectory data. We use weighting functions to incorporate spatial knowledge, such as segment lengths and turning angles, into our simplification algorithms. In addition, we measure the noise degree of a GPS point based on its spatio-temporal relationship to its neighbors. The effectiveness of our algorithms is comprehensively evaluated on real trajectory datasets with varying the noise levels and sampling rates. Our evaluation shows that under highly noisy conditions, our proposed algorithms considerably improve map matching accuracy and reduce computational costs compared to the state-of-the-art methods.