Ward, EJ; (2006) Urban movement - models of pedestrian activity. Doctoral thesis , UCL (University College London).

Preview

Text Ward_thesis.Redacted.pdf Download (118MB) | Preview

Abstract

As cities become more densely populated, there is increasing interest in predicting and understanding fine scale pedestrian activity to help plan and design more effective transport infrastructure. The research literature contains many pedestrian models which have been developed for specific applications such as emergency egress. However few studies provide generic modelling solutions which can be applied to transport interchange facilities or pedestrian use of urban spaces under normal conditions. This work examines the extent of pedestrian modelling as a whole, and then focuses upon two common types of pedestrian modelling at the macro and micro levels. The first model is the London Pedestrian Model, a static macro scale sketch plan method for the estimation of pedestrian occupancy. The basis for pedestrian sketch plan models are pedestrian counts and regression analysis which predict pedestrian flows as a function of key parameters derived from spatial and economic variables. Previous research such as that by Hillier (1993) in space syntax suggests that the configuration of the street grid (particularly the strategic lines of sight within the street network) can have a powerful influence on the distribution of aggregate pedestrian flows. In this study a range of spatial variables are tested, from the local measure of visibility (VGA analysis) to more global measures of network accessibility (derived from VGA analysis). Other variables that previous research has suggested are influential include land use and capacity and have been included in the London Model. Temporal and seasonal variables have also been included and found to be significant for the first time. The second model is JPed , a micro scale model capable of simulating pedestrians as individuals navigating local urban environments in real-time according to random, pre-planned or spontaneous trips to visit goals such as shops, restaurants, cash machines and transport interchanges. JPed simulations can encompass collision and pre-computed way finding algorithms as well as elements of pedestrian movement not typically found in current agent based models such as Legion (Connor 2004) and NOMAD (Hoogendoorn and Bovy, 2001). The model allows experiments to be run where pedestrians navigate through the environment using either route choices derived from direct sight information, previous experience in the form of memory maps or a combination of both tactics. Standard statistical tests of the output from the London Pedestrian Model and JPed gave correlations with observed data which equalled or exceeded previous models of pedestrian volume or movement. This finding proved pedestrian activity can be simulated at normal densities with some degree of accuracy using both simple statistical models and the agent based modelling paradigm.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item