An On-Line Arterial Route Travel Time Prediction Application Using ANFIS

Travel time study is basis to other traffic information service. Lots of factors like the intersection delay, the interference of non-motor vehicles and pedestrians affect the urban arterial traffic flow, making it displays much more complicated characteristics than the one of freeway. There are lots of efforts towards urban arterial route travel time forecasting methods; in this study, an ANFIS (Adaptive Neuro-Fuzzy Inference System) based real-time arterial route travel time prediction method is proposed, and tested using field data on arterial route segments in Shanghai, which covers both normal and failure conditions of detectors. Experiment results were then evaluated by a set of criteria. Results show that this approach has very good performance if being well trained with a large amount of data, even encountering incomplete information (detector failure), which validates the promising accuracy and robust of this approach. A sensitivity analysis of model inputs then carried out. Because the training procedure is usually costly, the direct city- wide implantation of this approach might not be feasible; however, with necessary improvement of training strategy, the proposed approach shall be even more satisfying.