A rolling-trained fuzzy neural network approach for freeway incident detection

Abstract

This paper develops a rolling-trained fuzzy neural network (RTFNN) approach for freeway incident detection. The core logic of this approach is to establish a fuzzy neural network and to update the network parameters in response to the prevailing traffic conditions through a rolling-trained procedure. The simulation results of some thirty-six incident scenarios in a two-lane freeway mainline case study show that the proposed RTFNN approach can improve the detection performance over the fuzzy neural network approach, which is based on the same network structure but without updating the parameters through a rolling-trained procedure. The highest detection rate is found at a rolling horizon of 45 minutes and a training sample size of 90 samples in this case study.