Research On Method Of Quantifying The Incident-Induced Delays On Freeway

The freeway system in China is in a stage of rapid development. However, with the significant increase in automobile ownership nationwide, freeway traffic volume is also growing rapidly and as a result the number of accidents and their severity have also increased significantly. Through quantifying the incident-induced delay on freeway in order to analyze the impact of incident is meaningful to developing understanding of the freeway operational condition and designing an effective incident management scheme to improve the safety and efficiency of facility.Based on the review of the state-of-the-art of domestic and international research, this thesis researches the theory and application of methods on analyzing the impact of incident on freeway. The main task is to quantify the incident-induced delay on freeway and the method of quantifying is that through comparing the real volume data with incident to predicted normal volume data and thus get the incident-induced delay. This method could eliminate the influence of recurrent factors (such as the holiday rush and daily rush hours) to total delay and just get the traffic delay caused by incidents. This could help formulae reasonable emergency measures when similar incidents happen. The main tasks are as follows:1. Data preprocessing. Based on the understanding that traffic flow monitoring system does not always produce reliable data, there may be some abnormal data in freeway traffic volume time series. Considering the distribution characteristics of traffic volume time series, this thesis proposed to use the method of wavelet transformation to identify and correct abnormal data. This method showed a good effect in an example.2. Volume forecasting. After preprocessing the initial volume data, we got normal traffic volume history data profile, which was prepared for the following prediction task. In this thesis we used RBF neural network to predict, with pre-incident normal volume history data as input, to forecast the would-be traffic condition had the incident not occurred.3. Quantifying the incident-induced delay on freeway based on the methods of data preprocessing and volume forecasting. Compare the predicted normal traffic flow data of downstream detector with the real traffic flow data of upstream detector, if the predicted data is larger than the real data, then we use the real traffic data of upstream detector and the real traffic flow data of downstream detector to estimat the incident-induced delay, orelse we use the predicted normal traffic flow data of downstream detector and its real traffic flow data to estimate the incident-induced delay. Through quantifying different types of incident-induced delay on different segments and in different times, we could know how much delay each type of incident might cause. This is helpful for guiding freeway operators to take reasonable measures when an incident occurs and thus improve operation efficiency. The quantified incident-induced delay could also be an important performance indicator for freeway incident management.