Advancement Through Research
Road incidents disrupt the flow of traffic and the cost of delays caused by an incident is huge, so to reduce the impact of an accident, traffic management centers must detect it quickly and remove it from the highway. Fast and efficient automatic incident detection has been a major goal of transportation research for many years, and many algorithms based on loop detector data have been developed and tested for Automatic Incident Detection (AID). Currently, research on incident detection algorithms is focused on artificial neural networks (ANNs) suitable for real-time detection, which are more reliable than statistical methods.
In this study, the fuzzy-based incident detection algorithm – FuzzyARTMAP - which is evaluated to be superior to the existing neural network models, was used for real-time accident detection on freeways and the reliability of the proposed model was verified using simulation traffic data. In particular, in this study, an optimal learning method was proposed to prevent overtraining and undertraining problems that occur in various neural network models, including FuzzyARTMAP, and to obtain optimal results.
Automatic Incident Detection (AID), Artificial Neural Networks (ANNs), FuzzyARTMAP, Optimal learning, Overtraining, Undertraining
