عنوان مقاله [English]
Public transportation systems such as bus and railway transportation play main role in transferring passengers and reduce problems in large cities due to low costs of travelling, high flexibility, ability in passengers transferring.
Increasing user’s satisfaction and optimal usage of existing facilities without planning is impossible. Vehicle assignment to routes, preparing time tables of trips, monitoring and implementation of schedules are the most important activities in public transportation planning.
Common process of planning in public transportation systems are including four steps: route design, generation timetables of trips, schedule of vehicles and schedule of crews.
In this process, short term activities at the beginning of each months or optimistically, at the beginning of each days, based on historical data set are planned. Hence, this type of planning is referred to as static planning.
In most transportation systems such as Tehran bus transportation system, static planning, in which plans are fixed in different months, is used. However, recently, Intelligent Transportation System (ITS) and Advanced Traffic Management System (ATMS) have been extended in public transportation systems. These systems prepare proper data set from location and state of vehicles and traffic variables and change static planning to dynamic planning.
Time of travelling is the most important factor in both of static and dynamic planning. Public transportation systems planning without any information about bus travel time between stations is impossible and lead to unreal plans.
In addition, estimation of travel time can be used for managing passenger’s demand and traffic via information systems as well as planning the other public transportation systems such as subway, taxi and so on.
In the previous researches, probability distributions are used to estimate the bus travel time in short period of time. Continuous changes of factors over travel time such as changing in traffic conditions in a route affect on probability distribution. Hence, a fixed probability distribution with known parameters cannot be considered for time travel variable during a day.
Development of proper forecasting models should be considered to estimate travel time in all time periods during a day precisely. In dynamic planning, one line data sets are available in addition to the historical data set available in static planning.
Forecasting methods with different performances can be designed and implemented to estimate travel time of public vehicles in both static and dynamic planning approach.
In this paper, proper variables are defined to measure the effect of time, traffic, and location factors on the travel time of public vehicles in a route. The effect of the mentioned variables is investigated on travel time of bus system in Tehran.
Two forecasting models are designed based on the identified variables to estimate travel time of public vehicle between two successive stations in a route.
First, a model based on historical data set and multiple linear regression is designed in which the relationship between travel time variable and the variables affect it is described.
Then, a feed forward artificial neural network is developed to estimate travel time based on one line and historical data set. This model can be used in dynamic planning approach.
Regression and artificial neural network models explain 75 percent and 94 percent of time travel variation, respectively, based on the data set from Tehran public transportation system.