Lossless Arithmetic Coding of Discrete Vehicular Trajectories

Tracking the movement of vehicles by transmitting or storing their position in regular intervals results in a great amount of redundancy, since vehicles are bound by physical barriers that greatly limit the possible movements. In a previous work, a lossy compression scheme was designed that exposed the redundancy in such discrete trajectories and provided an efficient arithmetic code for them. In this work, the principles of the lossy coder are translated to the lossless domain. This includes the analysis and solution of challenges introduced by the usage of a geodetic coordinate system, which is required to avoid information loss. We furthermore introduce different probability models for the arithmetic coder that cover different use cases and offer great flexibility in terms of resource consumption and required sample data vs. compression. These models are evaluated using real-world data from the OpenStreetMap project. The results show that the discrete trajectories encoded with our scheme are about 20% smaller than the encoding of the best existing lossy and lossless schemes.