Use of accelerations connected to GPS data to identify hazardous road locations

Hazardous road locations (HRL) are traditionally identified on the basis of reported accidents. However, the number of reported accidents is low and the dark figure is high. Hence identification of HRLs in traditional way is associated with uncertainty. Also, it is retrospective, i.e. the accidents have to occur before road safety enhancements can be made. The objective has been to build a predictive model for identification of HRL on the basis of Floating Car Data (FCD) associated with acceleration data. Strong decelerations are often involved in road accidents. From conflict studies it is known that there is a connection between the number of conflicts and the number of accidents. The hypothesis was that concentration of jerks, which is the derivative of the deceleration, can be used for identification of HRL and that these are better for this task than deceleration. A study is based on data from three months of driving by 380 cars, which is equivalent to a distance of 1.9 million km. This indicates that concentrations of decelerations are located where there are high traffic volumes, but not that the design and furniture of the road on the location can explain an HRL. This also indicates that concentrations of jerks can be found where the road has safety problems or is inappropriately designed. This is supported by literature dealing with this topic. On the other hand, jerks are very sensitive to unevenness in the road surface, and data from these locations must be removed in order to have reliable identification on the basis of jerks. It is therefore expected that focusing on position and acceleration changes 1-2 seconds before/after significant jerks will form a reliable basis for identifying HRLs.
The purpose of the abstract is partly to discuss the results so far and partly to discuss and explorer if these combinations of time series surrounding the jerks are a feasible approach to identify HRL.