Animal street crossing behavior
An in-depth field study for the identification of animal street crossing behaviour using the AIMATS-methodology
Based on the police recorded accident data in the German federal state of Saxony (2007-2014), 9.3 % (approx. 85,000) of all accidents involve animals. In 2015, 2,580 accidents involving animals caused injuries in Germany. In order to design ADAS (Advanced Driver Assistance System) in a way that helps to avoid such accidents, it is necessary to understand the animals' behavior. Current methods to observe animal behavior are using vehicle mounted NDS (Naturalistic Driving Study) data. This kind of NDS is expensive considering the number of relevant data sets recorded. This paper delivers the results of a one-year field study that used a new methodology based on in-situ recording units integrated in the infrastructure at critical sites. This way, vast data sets of animal street crossing scenarios can be generated in a quality similar to the one of NDS methods - yet at a relatively low cost. The definition of the scenarios is based on an in-depth investigation method which was presented at the ESAR conference (Hannover, Germany) in 2016 and is called ""AIMATS"". An accident data analysis of approx. 85,000 police recorded accidents with wild animal involvement in Germany made it possible to identify locations with a high possibility of accidents involving animals. These locations were observed by means of an infrared camera with a 50Hz frame rate. The recorded camera data allowed a detailed analysis of the movement of all road users. An automated analysis of the recorded results delivers typical and realistic models of the behavior of animals that have encounters with other road users. For this study, we assumed that the animal behavior at near miss scenarios is the same as their behavior in accident scenarios. This has been confirmed. This paper describes the results of a large-scale infrastructure-based traffic observation using the AIMATS methods. This method can be used for all traffic scenarios at a relatively low cost rate per scenario.