Experiments with tw-lines

This repository contains the data we collected while experimenting with tw-lines, along with the code to run the experiments.

Unfortunately, some of the data and code was lost. I tried to reconstruct the code, but if you try to execute it and run into errors, please contact me.

Data

We generated hundreds of artificial LinTim datasets, ran line planning algorithms on them and evaluated the results. The evaluations can be found in the folder collected_data. The different data tables correspond to different experiment setups. Our setups vary in how we generated the line planning instances and which algorithms we evaluated.

Running Experiments

Prerequisites

• LinTim (https://lintim.net/)
• tw-lines (https://github.com/urinstinkt/lptw)
• Python 3
• Install additional LinTim tools from the tools folder

Usage

experiments.py contains the main code to run our experiments. Before using it, you need to adjust the directory definitions at the top of the file.

At the bottom of the file, the different experiments we ran are defined. Currently experiment9 is enabled; if you want to run a different one, you need to edit the code.

The experiments are intended to run for a very long time, either until manually stopped by the user or until the generated instances become too large. There are two modes to execute experiments.py and usually you want to run them in sequence:

• "run experiment" with parameter -r: Generates LinTim datasets and evaluates them until stopped
• "collect experiment results" with parameter -c: Compiles all evaluation results from a previous run into a nice table, as seen in collected_data.