Kinematics-enabled lossless compression of freeway and arterial vehicle trajectories

Abstract

This paper shows how embedded instantaneous kinematic information from files of vehicle trajectories can be exploited to better enable data compression algorithms for typical files containing all such trajectories from a roadway segment over some period of time. Such files typically contain other relevant information, such as vehicle class, lane number, and so on, and effective ways of compressing these variables are demonstrated as well. The test files are taken from the Next Generation Simulation project, as those files are to date the de facto standard for large trajectory repositories suitable for studying traffic flow theory. The advent of connected vehicles suggests that the data collection, storage, and hence, compression needs in this arena will continue to grow. We develop compression algorithms that exploit collective vehicle kinematics of the first and second order to enable greater compression ratios. Using the context-free .zip and .7z compression routines as baselines, we compare three schemes. The first scheme treats each column independently and does not recognize any variables as kinematic and produces compression gains approaching 2:1. The second scheme also compresses each column independently, but recognizes that some of them contain kinematic variables, and hence relationships along the rows, resulting in gains of around 4:1. Finally, the scheme that uses all of this information, plus kinematic correlations across columns, produces gains on the order of 5:1.

Keywords:

• Data compression
• vehicle kinematics
• vehicle trajectories

Disclosure statement

No potential conflict of interest was reported by the author(s).