Optimizing Rail Transit Routes with Genetic Algorithms and Geographic Information System

Abstract

The location of rail transit lines and stations is a challenging problem because the number of possible alternatives is astronomical, the number of requirements and constraints is very large, and the evaluation of each alternative is based on numerous measures of effectiveness. Furthermore, large amounts of geographic information of many types are needed to compute some of the effectiveness measures. For instance, detailed spatial data on household locations and characteristics are needed to estimate accessibility to transit stations and the resulting access times and effects on demand. Detailed information on existing street networks, topography, and geology is needed to estimate construction costs and detailed information on land use and land values are needed to estimate right-of-way costs. In this paper we adapt the genetic algorithm and geographic information system approach to rail line optimization. Two types of costs are formulated and evaluated in the model: (1) operator cost which consists of track-related construction cost, station construction cost, right-of-way cost, earthwork cost, and rail operating cost; and (2) user cost which consists of access cost, riding time cost, and wait cost. The model is applied to a section of Anne Arundel County, Md., to reoptimize between a pair of stations an existing light rail transit line connecting Glen Burnie to the Baltimore Washington International Airport. The prototype model presented here produces very good alignments based on its criteria and can be extended to use far more comprehensive and complex evaluation functions. Additional factors and cost components that should be considered in more advanced models are also discussed.