Here, we implement a solver for the traveling salesman problem for a round trip and real-world data.
In the traveling salesman problem for a round trip we are given a list of locations. We want to take a round trip in which we start and end at the first location, and visit each of the other locations. We want to visit the locations in an order that minimizes the total distance traveled.
The user of our traveling salesman solver provides a list of addresses/location names. Using the openrouteservice.org API, we resolve the given names to locations on the world map. Subsequently, we use a genetic algorithm to determine the shortest round trip such that the beginning and end location are identical, and each location is visited.
To use the openrouteservice.org API, the user needs to open a free account at openstreetservice.org and create an api key. The example just below shows how the api key is loaded into the solver.
We provide a short example on how to determine and visualize the shortest round trip. The following example is from this notebook.
from traveling_salesman import traveling_salesman
# To resolve names to locations on the world map, and to obtain the traveling
# distances and routes between locations, we need to open a free account at
# openstreetservice.org and create an api key. Here we point to the text file
# that contains this api key:
parameters = {'api_key_filename':'api_key.txt'}
# Instantiate class
ts = traveling_salesman(parameters=parameters)
# Set locations
names = ['Lindwurmstr. 167 München',# salesman always starts at first location
'Museumsinsel 1, 80538 München',
'Sendlinger Tor München',
'Pinakothek der Moderne, München',
'Rotkreuzplatz München',
'Nordbad, München',
'Münchner Stadtbibliothek Maxvorstadt',
'Schellingstr. 4 Muenchen']
names_resolved, locations = ts.set_locations(names)
# To ensure that the provided names have been resolved correctly, one might
# want to compare names[i] with names_resolved[i]. The latter contains the
# name of the found location that will be used for the traveling salesman
# problem
# Solve the traveling salesman problem
results = ts.solve() # takes ~30s on my MacBook Pro
# by default, ts.solve() solves the traveling salesman problem three times
# and takes the best path of the three solutions
# Output:
# Found shortest path:
# 1. Lindwurmstraße 167, Munich, BY, Germany
# 1.74 km
# 2. Sendlinger-Tor-Platz, Munich, BY, Germany
# 1.94 km
# 3. Museumsinsel 1, Munich, BY, Germany
# 2.85 km
# 4. Schellingstraße 4, Munich, BY, Germany
# 0.89 km
# 5. Pinakothek der Moderne, Munich, BY, Germany
# 0.84 km
# 6. Münchner Stadtbibliothek Maxvorstadt, Munich, BY, Germany
# 1.36 km
# 7. Nordbad, Munich, BY, Germany
# 2.91 km
# 8. Rotkreuzplatz, Munich, BY, Germany
# 4.93 km
# 9. Lindwurmstraße 167, Munich, BY, Germany
#
# Total distance: 17.46 km
# Plot result
ts.plot_shortest_path(filename='munich_car.html')
# filename is optional. If no filename is provided, nothing will be savedHere is an illustration of the resulting round trip (locations are marked as blue dots, the routes between them as colored solid lines):
An interactive map is obtained by opening the file munich_car.html in a browser. This interactive map can be viewed directly here.
In the example notebook, also the shortest round trip for the same locations and a pedestrian is solved. With a total distance of 16.00 km, the pedestrian has a slightly shorter route:
Here is the corresponding interactive map.
To use the module simply clone the repository, move the file traveling_salesman.py into your working directory, and import it in your python script as shown in the above example (or the corresponding notebook).
To download the repository and install the dependencies (numpy, requests, plotly), run the commands
>> git clone https://github.com/juliankappler/traveling_salesman.git
>> cd traveling_salesman
>> pip install -r requirements.txt
>> pip install .