Variable Window Mechanism of Candidates for Map Matching

The ubiquitous popularization and deployment of mobile sensor technology has enabled GPS to observe and record vehicle states. Digital road maps as well as transport information of vehicles also has brought great convenience to intelligent transportation systems (ITS). However, raw GPS locations are usually too noisy and thus cannot be used directly by downstream applications. To alleviate the issue of location noise, map matching has been investigated to chronologically map a series of geographical GPS locations onto the actual road segments. In this paper, a variable window size for the candidate selection mechanism is introduced, which can confine the search region to optimize map matching process. We conducted experiments with a real-world road transport dataset and evaluated the performance to validate the effectiveness of the proposed method.