Generation of road maps from trajectories collected with smartphone - A method based on Genetic Algorithm

Smartphones and automotive GPS have considerably boosted the use of digital road maps. For this reason, they must be updated regularly with accurate new data. The methods currently used to generate maps - photogrammetry and collaborative editing - have low frequency of update because they depend on manual intervention. By using an automated method it should be possible to improve map update speeds while maintaining similar level of accuracy. The literature presents some approaches for automatic road map creation using moving objects, but none of them is prepared for continuous update. Therefore, this work aims to propose a new automated method that uses trajectories provided by GPS receivers integrated in smartphones. It is assumed that the points that represent the center of the roads can be found through approximations provided by Genetic Algorithm. After that, these points are combined to generate the road map. However, the use of trajectories collected with smartphones provides some challenges, such as: elimination of data with bad accuracy, identification of the means of transport used and reduction of the volume of data processed. Thus, the objective of this work is to propose a method that cleans, analyzes and enriches data from smartphones to generate accurate road maps that can be continuously updated, using Genetic Algorithm. Tests indicate that the proposed method can generate maps with quality similar to the reference maps with less than 2 m of difference in average. Additionally, a comparison between the Fuzzy C-Means algorithm and the Genetic Algorithm shows that the later is a little slower but generates more accurate results.