MARVin - Model for Assessing Risk of Road Infrastructure

Abstract: The project MARVin is about assessing the coherences between road infrastructure and road accidents. A large amount of data on the Austrian roads was gathered with the RoadSTAR (Road Surface Tester of arsenal research). This data was put into a database associated with accident data (about 500.000). The result was a database of about 21.500 km of road where all the road parameters belonging to a certain accident can be retrieved. MARVin is a practical tool to audit existing roads as well as planned roads and to develop accident preventive measures in terms of road construction and road infrastructure using a mathematical method called Dynamic Time Warping (DTW; see chapter "Similarity- Search"). One approach is to take a road segment where accidents happened and to find a very similar piece of road with roughly the same parameters. These road segments can be an interesting lead to reduce the number of accidents caused by road conditions. But it could also be the case that on a certain piece of road a lot of accidents occurs while on another road section, with very similar road parameters, no accidents happened at all. In this case there can be other parameters that influence the road safety such as e.g. speed limits or traffic density etc. Therefore MARVin is also planned as an open platform to integrate also relevant data like traffic flow data or weather data.