Estimation of Travel Time Based on Forecasted Precipitation

摘要

Currently, the collection of traffic information is increasing due to the expansion of ITS infrastructure construction. Moreover, a wide variety of traffic information is being provided for driver based on copiously collected traffic information. In the various studies regarding application of the traffic information, a study on traffic forecast is actively in progress. The traffic forecast can provide the accurate travel time to driver. In this paper, travel time estimation model was suggested by adopting forecasted precipitation data. The model was also based on traffic data collected by road-side equipment at Hanbat-daero, which is the primary section of arterial roads in Daejeon, Korea. Rainfall data was plugged into the artificial neural network for training in order to take account of the weather conditions that may possibly affect the traffic flow. In network training process, we selected the back propagation algorithm, and the model was constructed by testing how much sensitively react according to the change of rainfall. Mean Absolute Percentage Error and Root Mean Square Error were used for the reliability assessment of constructed model. In the results of the reliability assessment, the value of MAPE between the estimated travel time and observed travel time on a rainy day was 5.4013%, and RMSE was 0.37306. In addition, MAPE was 3.6539%, and RMSE was 0.21265 in sunny day. These results indicate the superiority of the traffic forecast.