Development of an automatic incident detection algorithm for freeway based on multi-level alarming system and artificial neural networks

A substantial reduction in the delay can be achieved by early detection of the incidents and prompt response to divert the traffic in the upstream flow. Since the late 1960s, Automatic Incident Detection Systems (AIDS) have been developed and implemented to help traffic management authorities. However, high false alarm rates and low poor performance have caused some authorities to abandon AIDS. To enhance the reliability, transferability and economization of AIDS, in this paper we design a framework for freeway AIDS with three-level alarm policy, and a AID algorithm based on the Artificial Neural Networks (ANN) technology that only need single detector. It was proved by simulated data that the model built on one segment can be used to other segments, and all three measurements (DR, FAR and MTTD) are superior to the objective algorithm. Furthermore, the total cost of freeway incident management system can be reduced due to only single detector is needed for one detector station.