gtfs-isochrone

create isochrone geojson from gtfs data

view demo

Usage

• locate a folder containing GTFS data as csv files. Required are:
  • stops.txt
  • calendar_dates.txt
  • trips.txt
  • stop_times.txt

git clone https://github.com/BenjaminHabert/gtfs-isochrone
cd gtfs-isochrone

# create python3 virtual environment
apt-get install build-essential python3-dev
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# download and extract data
mkdir -p data/orleans/
curl https://data.orleans-metropole.fr/api/datasets/1.0/donnees-du-reseau-tao-au-format-gtfs/alternative_exports/export_gtfs_du_29_06_au_30_08_20_zip > data/orleans.zip
unzip data/orleans.zip -d data/orleans/

# prepare the data (required only once)
python run.py prepare data/orleans/

# run the development server
python run.py server data/orleans/

# get geojson from a query on http://localhost:9090/isochrone
# example of valid query with this dataset :
# -> http://localhost:9090/isochrone?lat=47.9007&lon=1.9036&duration=60&start=2020-07-02T13:00:00

Deployment

Manual deployment on AWS EC2:

git checkout master
git pull
sudo ln -s /home/ubuntu/gtfs-isochrone/api.nginx.conf /etc/nginx/sites-available/apps/orleans-api.conf
sudo nginx -s reload
bash start_server.sh

Notes

Deployment notes

I use uwsgi to run the file server.py as explained here.

nohup uwsgi --socket 127.0.0.1:3031 --wsgi-file server.py --master --processes 3 > uwsgi.log &

I changed nginx configuration as follow (/etc/nginx/site-available/default):

server {


	location /gtfs-isochrone/ {
		# remove this part of the url
		rewrite ^/gtfs-isochrone(.*)$ $1 break;
		include uwsgi_params;
		uwsgi_pass 127.0.0.1:3031;
	}


    # ...
    # listen 443 ...
}

algorithm pseudo-code

PREPARE
- add date to stoptimes using trips + calandar_dates
- compute walk duration between all stops

INITIALIZE
- input: start lat, start lon, start-time, max duration -> end-time
- filter stoptimes between start-time and end-time
- filter: walk duration too long

START
- from starting position, get all reachable stops. Columns:
    - stop_id
    - arrival_time (ie start time + walking duration to the stop)

LOOP
- from selected stops -> find reachable stoptimes
    - same stop id
    - end-time > trip-stoptime > arrival_time
    - if no reachable stoptimes was found: exit loop
- expand reachable stoptimes
    - next stoptimes of the allready found trips
- split stoptimes:
    - A: those that were marked reachable
    - B: all others, kept for next loop
- from group A stoptimes:
    - drop duplicate stops: keep earliest arrival_time
- walk from the new stops to all possible stops -> new arrival times
    - remove if arrival time > end-time
    - drop duplicated stops: keep earliest arrival_time
- we now have a new list of reachable stops (stop_id, arrival_time)
- if it is the same as before: FINISHED
- else:
    - restart loop with these new selected stops


CONCLUDE
- we have list of (stop, shortest_arrival_time)
- create circle with radius = walking distance for each stop (shapely.buffer)
- add circle from start position
- merge all circles in a single shape (shapely.union)