From driving trajectories to driving paths: a survey on map-matching Algorithms

With the widespread deployment of built-in Global Positioning System (GPS) devices, numerous volumes of driving trajectories can be recorded conveniently. Affected by GPS equipment precision and driving environments, raw GPS trajectories will deviate from the paths that vehicles really drove on. Such inaccurate data is not beneficial to the upstream applications. Therefore, map matching is applied to identify the true driving paths in the road network from the GPS trajectories data. A lot of studies in the filed of map matching have been proposed, but there still exist three problems: (1) there lacks a comprehensive review on recent map matching algorithms with new techniques; (2) the existing map-matching algorithms still fail to meet the requirements of both high precision and high efficiency simultaneously; (3) there is a lack of comparison between various types of matching algorithms on a unified experimental environment. In this paper, we review the existing matching algorithms and propose a new categorisation based on their methodologies. The proposed categorisation can better reveal their properties and facilitate the future utilization. In addition, we conduct an experimental comparison among four representative algorithms to give a deep insight to the properties of different categories. Experimental results reveal the importance of some solutions to improve matching accuracy and efficiency.