Public Transit Stop Placer

Quinn Webster, Joe Bresee, Scott Garneau

This project explores how to place bus stops for a transit route in a way that improves rider access without making the stop spacing unrealistic. The goal is to find a stop layout that balances coverage, walking distance, transfer opportunities, and proximity to useful destinations.

The implementation uses a genetic algorithm. It starts from candidate stop locations along a route, evaluates each candidate layout with a weighted fitness function, and then iteratively evolves better stop sets over multiple generations.

What the project is trying to do

For a chosen transit route, the script aims to place stops that:

• improve access for nearby residents and equity-weighted populations
• keep stops reasonably spaced along the route
• support transfers to other routes
• stay close to important destinations such as schools, clinics, supermarkets, and community facilities

The result is a set of stop locations that can be viewed as an interactive map.

How it works

1. Load route and transit data from the project’s GTFS-style text files.
2. Generate initial candidate stop layouts along the route.
3. Score each layout using a custom fitness function based on several weighted criteria.
4. Apply genetic algorithm steps:
   • select strong parent layouts
   • combine them through crossover
   • mutate some children to explore new stop positions
5. Keep the best-performing layout and save it as an HTML map.

Project structure

• code/ga.py — main genetic algorithm workflow
• code/weight_function.py — fitness scoring and spatial evaluation logic
• code/config.py — GA parameters and route settings
• code/busRoutes.py — route and stop generation helpers
• code/evenlySpacedBusStops.py — candidate stop generation logic
• code/plot_actual_stops.py and code/best_bus_stops.html — map-related output examples

Setup

From the project root, install the required Python packages:

pip install -r requirements.txt

If you run into missing geospatial packages, install them as well:

pip install geopandas osmnx shapely

Run the optimizer

The script expects to be run from the code directory because it reads data files relative to that location:

cd code
python ga.py

This will print generation-by-generation progress and save the best result to:

code/best_bus_stops.html

Notes

• The current configuration targets route 95 by default. You can change this in code/config.py.
• The fitness function is intentionally weighted and can be adjusted depending on what matters most for the route being studied.