Traditional concepts of accident prediction models and safety performance functions are based on certain input values derived from infrastructure and accident occurrence. Usually, driving behaviour is not taken into account even though its impact is known. So far, an implementation of driving behaviour in APM fails due to missing valid behaviour data. Collecting driver and driving behaviour data is time consuming and expensive. Therefore, often, only small parts of road networks are investigated or where a small number of samples are available. Nevertheless, in order to understand accident occurrence better and more accurately it is necessary to develop a scientific methodology that considers at least these aspects of driving behaviour, which already has been investigated and for which appropriate models have been derived. This is the case for speed behaviour which was shown in numerous research studies. Within the framework of work package 10 of the RIPCORD-ISEREST project an approach was developed that considers speed behaviour together with infrastructure parameters. In contrast to other approaches, driving behaviour was not implemented as an additional model factor, but was rather used to classify the alignment based on its impact on speed. The results have shown that the implementation of speed behaviour is useful and improves models to predict accident occurrence. The approach of RIPCORD-ISEREST has been taken up again and was further developed by applying an improved speed prediction model. This model was used to detect driving behaviour related geometric elements and element combinations for the analysis of accident data. (Author/publisher)
Abstract