In order to overcome the subjectivity of existing pedestrian route choice models, an alternative choice model is presented based on the utility equation. It is composed of several indirectly objective characteristic variables, including the height, length, and width of interlayer facilities; speed of automated facilities; and carry-on luggage. Considering the scene that pedestrians choose between the stairs or escalators, an extended binary logit model is developed. Calibration and validation of the model are accomplished by using the data collected in four typical passenger transfer stations in Beijing, China. The results show that the proposed model has an average accuracy of 86.56% in bidirection for predicting pedestrians’ behavior. An interesting phenomenon can be found that the length of facility has poorer impact than height on pedestrians’ route choice behavior. Some quantitative and irradiative conclusions have been illustrated on the relationship between the selection probability and the variables, which is expected to be valuable for extracting the implicit theoretical mechanism of passenger choice behavior.
Integrated development of the modern passenger transfer hubs realizes the combination of intercity and in city transportation within or in the same space layout. Centralized transferring of pedestrians greatly improve the operating efficiency of the urban passenger transport network. Passengers transfer space has a multilayer and attaches a certain amount of interlayer facilities connection, such as stairs, escalators, elevators and slopes. The three-dimensional structure increases the complexity of the traveler’s behavior and activities in the hub. At peak time, due to the impact of a large number of passengers, bottlenecks usually appear at interlayer facilities, which induces local congestions and unnecessary delays. Therefore, understanding of the role of the relationship between these facilities and pedestrian behavior has become a very compelling issue [
As is known to all, in the process of pedestrians through level bridging facilities, the physical exertion must be paid by the tourists that have high-positive correlation with the height of facilities, which plays a key role in their choice decisions, especially those carrying luggage. Moreover, this must be related to the fundamental structure of facilities and current environmental factors such as flow density. There are lots of researches on modeling and analyzing the passenger choice behavior of facilities. Daamen et al. [
In short, these studies have focused on the relationship between passenger flow and interlayer facilities to reveal the theoretical mechanism of passenger choice behavior, ignoring the effect of interlayer facilities physical properties on passengers. As for the models, discrete utility theory has been extensively used on these issues, whether in horizontal or vertical direction. The difference between these models is the selection of influencing factors on calculation utility function, such as the generalized consumption [
On the basis of previous research, the height of facilities was proposed as an independent factor in a utility model to describe the relationship between height of facilities and pedestrian facilities choice behavior.
After a brief literature review, the new model was put forward, and some of the features were expounded in detail. Then, the scene of the traffic data which model calibration required was introduced, and the model calibration results were shown. Finally, theoretical characteristics were illustrated which extracted from further analysis of the model and came to some meaningful conclusions.
To calculate the utility function, we need select the influencing factors. These is the most fundamental problem. An extended literature study on empirical data and modeling of pedestrian route choice can be found in [
Passengers will take the corresponding physical strength when they are using level bridging facilities. Within reasonable delay time, people tend to choose less expensive physical facilities. That is to say, physical consumption plays a substantial role in pedestrians’ choice in vertical dimension. If passengers carrying luggage (we consider only the heavy luggage without light handbag), more energy will be consumed when passengers go through the non-automated facilities.
After an overview of the influencing factors above, the structure and basic consumption are selected for subsequent modeling; passenger property will not be considered. Even, the items of data collection are corresponding with them.
Theory of disaggregate model is based on the hypothesis that consumers choose to pursue utility maximization. In the issue of the passenger facilities selection, passenger choice behavior and consumer have the same principle. Logit model, which has advantages of simple structure and strong applicability, divides utility into uncertainty utility and random utility and assumes that the random utility obeys certain probability distribution, thus obtains the probability of travelers choosing different transport facilities.
Assume that the passengers can select the options independently, the utility model
Constructing and evaluating the utility function is a key link in the process of correctly analyzing passenger choice behavior. Borgers and Timmermans [
Combining with the characteristics of passenger choice behavior in vertical dimension, the time and physical consumption are select and used as characteristic variables, as shown in Table
Characteristic variables.
Characteristic variables | Variables represent | |
---|---|---|
Escalator | Stairs | |
Time consumption | ||
Walking time ( |
|
|
Delay time ( |
|
|
Physical consumption | ||
Height ( |
0 |
|
Luggage ( |
0 |
|
Then, the new
Set
To exclude the interference of other transportation modes and obtain pedestrian traffic flow parameters in a range as wide as possible, four transfer stations were selected as the observation sites in Beijing, such as Xizhimen subway station, Beijing South Railway Station, Zoo Station, and Zhichun Road Station, for they have large amounts of vertical pedestrian travel demand. And, the number of different choice situations and route alternatives is reasonably high in these stations.
The way of basic data acquisition is video observations, as described by Cheung and lam [
Video snapshots.
In ascending
In descending
The factors considered as well as the corresponding data obtained from the videos in the study include date, peak time, station, direction, structure of facilities (height, length, and width), speed of automated facilities, the number of people in queue for escalator, and whether carry-on luggage.
Model calibration is divided into three steps. Firstly, the form of the utility function and the characteristic variable must be determined, on behalf of calibration of the model. Secondly, using the maximum likelihood estimation method to calibrate the parameters and making some preparations for model validations. Finally, the covariance matrix is used to make
Calibration result.
Direction | Type |
|
|
|
|
|
---|---|---|---|---|---|---|
Ascending | Value | 6.6324 | −0.2501 | −0.5986 | 0.8642 | 0.9976 |
|
2.16 | 0.51 | 2.53 | 2.86 | −3.76 | |
|
||||||
Accuracy | 89.41% | Goodness | 0.86 | Conformity | 0.81 | |
|
||||||
Descending | Value | 5.9077 | −0.5539 | −0.7086 | 0.7331 | 0.8244 |
|
2.26 | −0.84 | −4.15 | 2.17 | 3.93 | |
|
||||||
Accuracy | 83.70% | Goodness | 0.79 | Conformity | 0.76 |
As the absolute value of the
In ascending direction
In descending direction
This section presents an examination of various properties of pedestrian choice behavior indicated in the percentages of the stairs that will be used. In line with [
Figure
Percentage of pedestrians using stairs by different waiting time.
When height is 5 meters and delay time is 20 seconds, the probabilities of each alternative are chosen almost the same. Once the balance is broken, passengers will more clearly to make a decision. In case of ignoring waiting time, the vast majority of people will not pay attention to the stairs. Escalator plays an important role in the process of transition between the layers. As the delay time changes from 10 to 20 seconds or more, the concern of pedestrians shows a dramatic shift, so much as more than 98 percentages of people choose stairs when the waiting time is for 30 seconds. Under the condition that the floor is not too high, it is necessary to appropriately increase the stairs width to encourage passengers to get through as soon as possible. Otherwise, a lot of passengers were forced to crowd in front of the facilities, although they are willing to pay extra strength in climbing the stairs.
Figure
Percentage of pedestrians using stairs by different height.
For example, as the height reaches 20 meters, the traveler’s acceptable delay time will not exceed 35 seconds. Otherwise, a significantly increase occurs in probability of passengers choosing stairs. Then, a larger number of passengers will gather in the access corridor to the escalator.
In the design stage of infrastructures, corresponding to the height of the interlayer facilities, the area of access facilities to level bridging facilities must be large enough, at least for satisfying the demands of the corresponding number of passengers.
Figure
Effect of luggage on probability of passenger using stairs.
According to the analysis on effect of luggage to pedestrian’s choice behavior, a good proposal could be put forward. Compared to commuter staff, the proportion of intercity travelers which carrying luggage is higher and the number of escalator in intercity transportation hub should be more than transfer hubs in city, in order to improve the capacity of the automation facilities and provide better service to passengers.
This study developed a binary logit model with an alternative utility function to predict pedestrian’s choice behavior between stair and escalator in public transfer stations. Different from the other models, the generalized extreme value indicated by subjective perception of time consumption of passenger uses level bridging facilities that have been divided into two parts. One is time consumption which consists of walking time and delay time, and the other is physical consumption which is represented by the height of facilities and whether carrying-on luggage. In addition, ascending and descending data for calibrating and testing the function has been collected in five passenger transfer stations in Beijing, China, including ten stairs and escalator parallel scenarios.
According to the results, the average accuracy of 86.56% is extraordinary high for this type of model. The model also shows that the length of facility has little influence on passenger selection behavior. So, it is reasonable that the height is identified as an independent characteristic variable. Some phenomena have been observed with further analysis of the model. Passengers are more sensitive to facilities height in ascending than in descending. Due to the objectivity of the selected parameter, we can intuitively understand the relationship between pedestrian choice behavior in vertical dimension and their specific environment. Also, the estimated values for infrastructure types may directly be applied in a simulation tool, even as a theoretical reference for facility design decisions.
The authors declare that there is no conflict of interests regarding the publication of this paper.
This work was supported by the Independent Research Topic of State Key Laboratory (RCS2011ZT002), National Nature Science Foundation of China (71171015), and National High Technology Research and Development Program (2011AA110506).