Optimization of Spacing of Transit Stops on a Realistic Street Network

Abstract

A discrete model of bus stop location in which candidate stops are either selected or not has several practical advantages over classical continuum models. An evaluation method for stop sets that uses parcels as units of demand and the street network to model walking paths between transit stops and parcels has been proved effective and realistic. In this framework, the on–off counts at existing stops are used to allocate demand to the parcels in each stop's service area in proportion to the stops' trip-generating ability. The result is a demand distribution that matches existing counts and reflects variations in land use. However, with demand modeled on the street network, the placement of service boundaries midway between neighboring stops becomes invalid because of irregularities in the network of access streets and curves in the transit route. The dependence of a stop on more than its immediate neighbors for determination of its service area complicates the process of optimization of stop locations by use of dynamic programming. The proposed solution expands the state space so that a stop's service area is dependent on the two prior and the two succeeding stops. The resulting dynamic programming model was tested on two bus routes and found solutions that were better than the existing stop set and the stop sets proposed by consultants by use of simple yet state-of-the-art models. This paper describes a method for optimization of stop locations on an existing route that includes realistic and localized estimates of its impacts on walking and riding times and operating cost.