The Fuzzy Inference on Wet-Pavement Related Crashes

The occurrence of a wet-pavement crash is actually a complex system, and to a large degree nonlinear, as the behavior of drivers and the interaction between roadway and vehicle characteristics tend to be nonlinear. As a result, prediction of wet-pavement crashes is difficult to model mathematically. Fuzzy logic models, however, can model nonlinear functions of enormous complexity, and can be built on the experience of experts. At the same time we may easily express the operations of this complex system in linguistic rules using fuzzy variables. The developed fuzzy logic model uses four input variables: skid number, speed differential, traffic volume, and driving difficulty to predict wet-pavement crashes. The results of the model indicate that the fuzzy logic model performs very well, even with limited and imprecise data. Such a prototype can evolve into a tool for identifying sections of roadway that are likely to have more wet-pavement crashes than normal and to make safety improvements accordingly. It can also be used in planning and programming for testing the cost-effectiveness of wet-pavement crashreduction projects.