Summary |
In recent years, travel time information has received great attention, providing safety guarantees for travelers to choose suitable travel routes and improve travel efficiency. It is also an important part of urban traffic planning, traffic management and traffic control. Travelers need reliable transportation systems to plan trips and arrive at their destinations on time. And travelers not only care about the length of travel time but also want the travel time to be reliable. Therefore, in addition to providing a predicted value (such as mean, median), providing an estimated time range can give travelers more useful information. However, the time range provided by the existing travel time prediction range is too large, and the distribution within this travel time prediction range is unknown. The travel time reliability is the certainty and predictability of the travel time of the transportation system. Therefore, this study provides more travel time related information for travel planning based on the travel time reliability index.
In the past travel time prediction research, the prediction model part focused on combining the time dimension and the space dimension prediction model, so as to incorporate the time-dependent and spatial-dependent travel time influencing factors to improve the accuracy and efficiency; the data part discussed more information Sources of outliers due to detectors, data processing, etc. At the same time, the data used by the past travel time prediction model is mostly link traffic data, which is calculated by the difference in time, traffic flow and distance between the two detectors, and the result is less reliable for long-distance travel.
The main purpose of this study is to explore the travel time prediction and reliability, using the M06A data of Taiwan freeway for discussion. The data includes the time information of each trip through the starting point gantry and the ending point gantry, and the time between each gantry in the starting and ending points. The travel time prediction and reliability were discussed based on the preprocessed data of the research scope data and the Long-Short Term Memory (LSTM) prediction model of travel time in this study. The results show that the use of route screening and road segment travel time screening methods can improve the accuracy and reliability of travel time prediction. At the same time, the results show that the travel time within the scope of this study is more reliable during holidays and Taiwan summer vacations in June, July and August. From the perspective of policy planning, travelers during these times have high flexibility in choosing the starting time. If accurate and reliable travel time forecasts can be provided, travelers will choose different starting times according to their personal needs, which can indirectly reduce the total travel time of the system. |