Institute of Transportation, MOTC

       Walkability is the most essential part in developing sustainable transportation systems, since walking is the first and last mile of all green transportation modes. Without a good walking environment, it is impossible to make transportation systems greener. However, due to the budget constraint, the improvement of sidewalk facilities should be prioritized according to their walking demand. Based on this, this study aims to develop pedestrian flow estimation models of sidewalks under various sidewalk geometrics, built environment, and social economics/demographics so as to evaluate the level of service of the sidewalks as well as their improvement priorities.
        To do so, this study attempts to estimate the peak-hour pedestrian flow of sidewalks based on their geometrics and neighboring built and social economic environments by using regression model and artificial neural network (ANN). The pedestrian flow data were collected from the CCTV of 96 sidewalks under weekday and weekend, receptively, making a total of 192 samples. The morning and evening four peak hours were identified from a 24-hours video counting. A total of 27 explanatory variables were also collected, which are 7 sidewalk geometrics and facilities, 7 social economic and demographics , as well as 13 built environmental variables, including 6 land use variables, 3 POI (place of interest), and 4 public transportation. The estimated double-log regression model for weekday and weekend pedestrian flows shows 15 and 10 variables are significantly tested with adjusted-R2 of 0.6838 and 0.7425 and MAPE 7.18% and 9.60%, respectively. The tuned ANN models (Back-propagation neural network, BPN) with two layers show the MAPE of 8.28% (weekday) and 7.90%. Both regression and BPN models perform very satisfactorily.
       To show the applicability of the estimated models, this study further develops an evaluation framework by using Analytic Hierarchy Process (AHP) and interviewing 6 experts. The most important criteria are the effective width of sidewalks (0.1879), followed by signal control facilities (0.1707) and pavement damage (0.1707). A case study of sidewalks longer than 300 meters in Taipei city is then conducted. The distribution of these sidewalks on a two-dimension figure (pedestrian flow vs. LOS scores) is helpful to prioritize their improvement.