Improving GPS-based road network constructions in a post-processing step of global optimization

Since digital maps of high spatial and temporal resolution and precision are widely regarded as key enabler for advanced driver assistance systems and automated driving, the research field of automated map generation from crowd-sourced fleet data has gained importance in recent years. This paper proposes a novel optimization method for improving the results of known road-network construction algorithms in a post-processing step. The method is proven to be feasible by an evaluation of the results on real-world datasets. Furthermore, the applicability to traditional GPS data is explored.