A Cluster Algorithm for Road Surface Condition Data Constrained by a Road Network

Up-to-date maps of road surface conditions are essential for future vehicles and pavement management systems to ensure driving safety and comfort. There are different approaches to measure road properties like obstacles or unevenness automatically and to share these information by a global network. This paper presents a workflow to cope with complex cluster shapes in a road network. Therefore, the analysis is performed at first on each single road segment (RS) and finally on contiguous RSs. The hypotheses obtained by clustering have to pass a series of verification steps to gain trust. For this purpose, a statistical analysis and a comparison of different alternative hypotheses are used to identify ambiguities in analysis results. The improvements achieved by this will be shown for an example with measurements for several obstacles at an intersection.