Analysis of Pedestrian Level of Service and Capacity at Various Transportation Facilities

Abstract

Over the years, pedestrian mode is the most accessible mode of transportation, in which a person has freedom to move from one place to another easily and flexibly between different regions. Hence, sustainability of community development largely depends upon the establishment of efficient transportation facilities to serve pedestrian mode effectively in a transportation network. The review of literature suggests that the service quality of transportation facilities has been evaluated since nineties. However, the developments of transportation facilities in developing countries like India is still far away from satisfactory than the developed countries, because of the fact that application of pedestrian level of service criteria developed for homogeneous traffic conditions will fail to quantify the service provision criteria and heterogeneity of transportation facilities in developing countries. At the same time, handling the pedestrian traffic becomes extremely challenging due to limited space and resources. In accordance with this, pedestrian level of service, which is a complex expression represents the operating condition of transportation facilities and satisfaction levels of pedestrian experience while using these facilities. In this regard, this study aims at defining Pedestrian level of service categories for three transportation facilities such as railway foot over bridge, signalized intersection crosswalk and un-signalized midblock crosswalk by using qualitative data sets under non-uniform pedestrian flow condition. On the other part of this study, while going through the literature, various researchers developed several methods for capacity estimation. Also, Highway Capacity Manual (2010) provides procedures in detail for determining capacity of facilities, but due to the heterogeneous conditions of traffic flow in India, these models doesn’t fit well in Indian condition. Therefore, in the present study, Artificial Intelligence based Genetic Programming models, which are empirical in nature, are developed for railway foot over-bridges three facilities such as staircase, deck and junction by considering capacity as the dependent variable and geometric variables as independent variables, under non-uniform pedestrian flow condition in India. To achieve the objectives of this research, a broad spectrum of pedestrian perception data sets are collected from three transportation facilities: Railway foot over-bridges, signalized intersection crosswalks and un-signalized midblock crosswalks. Around, 12 railway stations’ foot over-bridges, 9 cities’ signalized intersection crosswalks and un-signalized midblock crosswalks were investigated in this perception study. From each investigated station or city, 600 pedestrian’s real time sense of satisfaction data are collected with the help of prepared inventory questionnaire. For capacity estimation of these railway stations’ foot over-bridges three facilities, namely staircase, deck and junction, pedestrian flow and inventory data were collected. With the help of videography survey, pedestrian flow data are collected. Consequently, by eye observations and field measurements of various geometrical variables, inventory data were collected. The method of sensitivity analysis is employed to determine the impact of an independent variable with a particular dependent variable under a given set of assumption. Factors sensitivity analysis for three transportation facilities were carried out in this study. Significant factors are identified to have influence on pedestrians’ satisfaction level at railway stations’ foot over bridges, such as: width, length, rise, trade, railing and surface under staircase, frequency and size are under landing, railing, surface, position with respect to halting of train and position with respect to entry/exit of station are under foot over bridge, display, announcement at proper time and at all the platforms are under information provided, and requirement of foot over bridge, escalator, travellator, more platforms and directional separator are under requirement of facilities. Similarly, location of signalized intersection, visibility of zebra crossing line, width of zebra crossing line, observation of signal color, pedestrian traffic signal color, time allocation for pedestrian crossing, location of zebra crossing, following of traffic signal, waiting time, vehicle occupy the zebra line, lighting of zebra lines at night, grade separation and requirement of subway and foot over bridge are the factors, which have significant influence on pedestrians’ service quality at signalized intersection crosswalks. For un-signalized midblock crosswalks, pedestrian crossing of median, median space, road surface and lighting system at night, safety in front of vehicles, time taken to cross the road, safety while waiting across the road, vehicle speed variance, nature of traffic volume, median opening, zebra crossing, choice between foot over bridge or subway facilities, requirement of zebra crossing, foot over bridge and subway, and lighting of zebra crossing at night are found to have significant influence on pedestrian’s comfort level under non-uniform pedestrian flow condition. Consequently, sensitivity of each railway stations’ foot over-bridges, cities’ signalized intersection crosswalks, and cities’ un-signalized midblock crosswalks are obtained. Defining the pedestrian level of service criteria for urban pedestrian facilities is basically classification problem and cluster analysis is found to be more suitable technique for solving this problem. In this study, k-means clustering technique is used to determine the Pedestrian level of service of three transportation facilities. In accordance with the HCM (2010), to define pedestrian level of service for these facilities, sensitivity is employed as service measure. An associative Pedestrian level of service category is assigned to six classes (A-F) for corresponding facilities. For railway stations’ foot over-bridges, pedestrian level of service categories were defined. Consequently, with respect to service levels, the railway stations are classified in decreasing order. More than 70% of the studied railway stations’ foot over-bridges are examined to offer pedestrian level of service ‘C’ or above. Similarly, pedestrian level of service classification for various cities’ signalized intersection crosswalks and un-signalized midblock crosswalks are determined. Accordingly, the cities’ facilities are categorized in decreasing order corresponding to levels of service. More than 65% of the cities’ signalized intersection crosswalks and un-signalized midblock crosswalks were found to be offering pedestrian level of service ‘C’ or above. Based on the obtaining offered service quality of these transportation facilities, which will help the transport authorities to identify operational issues of existing facilities, and to design a pedestrians’ friendly transport system with better pedestrian movement. Also, this will help in defining service levels for transportation facilities in other developing countries having non-uniform pedestrian flow condition like India. As defined by highway capacity manual (2010), capacity is the maximum flow that could be carried in the approach stream under prevailing geometric and traffic conditions. Various researchers over the years, developed several methods for capacity estimation. The traffic condition in India is very heterogeneous in nature, whereas, the traffic condition of developed countries is homogeneous in nature. Therefore, the methods recommended by highway capacity manual (2010), for estimation of capacity in terms of models under homogeneous traffic flow condition, which doesn’t fit well in Indian condition. Hence, Genetic programming modelling based on Artificial Intelligence is used for model development in the present study, as it’s the latest technique. In this regard, for railway foot over bridges’ three facilities, namely staircase, deck and junction, Genetic programming models were developed under non-uniform pedestrian flow condition in India. Models are developed for these facilities by taking geometric variables as independent variables and capacity as dependent variable. For development of three models, from total data sets, 70% and 30% of the data sets were used for training and testing of the models, respectively. Also, these models are obtained for a maximum of 50 generations with mutation probability of 15% and population size of 1000. For model development of staircase, in both training and testing stage, the statistical factors such as coefficient of determination of the model, Average absolute error, Mean absolute error, Root mean squared error were measured to be (0.92, 0.09, 0.12, 0.10) and (0.91, 0.11, 0.15, 0.13), respectively, which indicates that the model’s prediction capability is significant under non-uniform pedestrian flow. The method of data splitting was employed for the model validation and found that observed capacity varying 8% corresponding to predicted capacity with the measured R square value of 0.92. This shows that the capacity model for staircase found to best fit at 92% confidence level. In deck model development, the statistical parameters like R square of the model, Average absolute error, Mean absolute error, Root mean square error are measured to be (0.91, 0.10, 0.16, 0.12) and (0.88, 0.13, 0.18, 0.15), in both training and testing stage of the model development, respectively, which shows that under non-uniform pedestrian flow, the prediction capability of model is significant. Data splitting method was used for validation of model and found that the capacity model for deck is well fit at 91% confidence level. Also, the variation between observed capacity and predicted capacity is 9%, with the value of R Square is measured to be 0.91. Model development for junction, the prediction capability of the model is significant under non-uniform pedestrian flow as the statistical parameters such as R square of the model, Average absolute error, mean absolute error, root mean square error are measured to be (0.87, 0.12, 0.16, 0.14) and (0.84, 0.15, 0.19, 0.17) in both training and testing stage of the model development, respectively. For the model validation, method of data splitting was employed and found that observed capacity varying 13% corresponding to predicted capacity with the measured R square value of 0.87. Further, this signifies that at 87% confidence level, the capacity model for junction is best fit. The proposed models are suggested for application in the field by practitioners because the models provides unique equations for prediction. The models can be widely acceptable for their simplicity and manageability. Therefore, these models are quite useful for planning and design of respective facilities. Also, these findings will be useful for the researchers, planners and designers for making key policy decisions.