A Method for microscopic modeling of pedestrian delay at signalized crossings

Abstract

When assessing the pedestrian level of service at signalized crossings, an important measure of effectiveness (MOE) is delay, which may be affected by several factors, both human (e.g. age an d gender) or local (e.g. vehicular flow and signal timing )). The violation of the don’t walk indication affects significantly delay, as pedestrians accept gaps on traffic flow to reduce their delay. An opportunist is a pedestrian who searches for gaps durin g red , and modeling the opportunist behavior with precision is very important for modeling delay. One essential parameter regarding opportunistic pedestrians that needs to be estimated is the critical gap. When using microsimulation tools, such as Vissim a nd its Social Force and priority rules model s , it is essential to satisfactorily estimate the behavioral parameters used. Since there is still a scarcity of scientific works towards the use of microsimulation to model pedestrian delay, this M.S. thesis has as main objective the proposal of a method for microscopic modeling pedestrian delay at signalized crossings using Vissim. Four pedestrian crossings with different characteristics, such as number of lanes to cross an d vehicular flow, were analyzed. The violation rates observed were : 59% for Crossing 1 10% for Crossing 2 , 37% for Crossing 3 and 47% for Crossing 4. T he differen t violation rates are probably due to the availability of gaps (Crossing 2 has a higher vehic ular flow and, consequently, less adequate gaps , for example ). All pedestrians were treated as opportunists in the simulation, due to the difficulty of estimating the proportion of opportunists for both real and hypothetical situations. Vehicle arrival pa tterns were satisfactorily matched by making some adjustments to the network, such as modifications on car following parameters. Three methods for estimating critical gap were applied: the HCM’s, Chandra’s and Raff’s. A fourth method using microsimulation for estimating critical gap was proposed; the targets were the average delays per pedestrian type (man, woman, young, senior). The fourth method yielded the best results when comparing the methods in terms of the estimation of delay every 1 5 minutes. Mean absolute percentage errors of 19%, 10%, 22% and 54% were obtained for Crossings 1, 2, 3 and 4, respectively when considering pedestrians who arrived on red. The worst estimations happened for Crossings 3 and 4 due to delay peaks observed in two of the inte rvals ; the peaks were associated with high vehicular flows and low violation rate s Finally, a sensitivity analys is revealed that none of the Social Force model’s parameters had a significant impact on the estimation of pedestrian delay at signalized crossings.