social-transit-example

A script to procedurally generate small-scale example networks for public transit design with social access objectives.

This is part of my research project related to public transit design with social access objectives (see project home page here), which involves a case study of the Chicago bus network and primary care access. The purpose of this program is to generate small-scale synthetic networks for use in testing the solution algorithm, as well as for adjusting the search algorithm's tuning parameters and conducting sensitivity analysis.

I would not expect this program to be of use to anyone outside of my research group, but it is being made available if anyone is interested.

Summary

This program is meant to procedurally generate test instances which can be quickly evaluated by eye to decide whether they are "interesting". Calculating the operator and user costs of the network is expensive, so we will consider only the objective value as a function of the line fleet sizes. An "interesting" example should have a "rough" objective function which presents a challenge for the solution algorithm. A quick way to test this would be to choose a few random pairs of decision variables and vary each pair to generate tables of objective values. We would like for all of these tables to show non-monotonic, non-convex behavior.

The procedural generation should rely on a set of input parameters which can be quickly changed by the user. Based on these parameters, a random problem instance should be generated along with some metrics for evaluating whether the instance is interesting. There should also be some functions for quickly generating output files based on the current example network.

For simplicity, our generated network is based on a square grid made up of vertical and horizontal transit lines. A stop exists at each grid intersection as well as intermittently along each line. Population centers and primary care facilities are randomly distributed within the rectangular grid area, with population centers being somewhat evenly spaced and facilities being completely random.

Output Format

See format.txt for a full description of the output file format. In particular we are interested in outputting the following:

• arc_data.txt: arc ID, arc type, arc line ID, arc tail node ID, arc head node ID, arc cost
• node_data.txt: node ID, node name, node type, node line ID, node value (population or facilty weight)
• od_data.txt: unique ID, origin node ID, destination node ID, travel volume
• transit_data.txt: line ID, line name, vehicle type, initial fleet size, total circuit time, scaling factor, fleet lower bound, fleet upper bound, boarding fare, initial line frequency, initial line capacity
• vehicle_data.txt: vehicle type ID, vehicle type name, vehicle type bound, vehicle type capacity, vehicle operating cost

User Cost Search

Also included in this repository is a simplified version of the main solution algorithm from social-transit-solver. It has been modified to optimize the user cost rather than the social access objective, while ignoring the user cost constraint. It also consists purely of a local search rather than a hybrid tabu search/simulated annealing algorithm. This can be used to further refine the initial solution vector produced by the Mathematica script by modifying the initial fleet sizes to achieve lower user costs.