抄録
This paper proposes a new incident detection approach which can detect the incident before it occurs. The new incident detection model consists of a prediction mechanism for driving behavior by involving a car-following model to predict the car-following behavior. The model is then Predictable Incident Detection (PID) model. The test and development of the model used the simulated data. For checking the incident occurrence, a new parameter - distance gap between two following vehicles was adopted. Using artificial neural network technique with the parameter of distance gap between vehicles, the incident detection model was established. The test results indicate that the new method can actually predict an incident, i.e. with a negative time-to-detect. The proposed method can be a useful method when combined with other conventional methods to enhance the performance of existing incident detection systems.
