pyvroom 1.15.0rc3

Vehicle routing open-source optimization machine (VROOM)

Python Vehicle Routing Open-source Optimization Machine

Good solution, fast… in Python.

Pyvroom is an Python wrapper to the excellent VROOM optimization engine for solving vehicle routing problems.

The library aims to solve several well-known types of vehicle routing problems, including:

• Travelling salesman.

• Capacitated vehicle routing.

• Routing with time windows.

• Multi-depot heterogeneous vehicle.

• Pickup-and-delivery.

VROOM can also solve any mix of the above problem types.

Basic usage

>>> import vroom

>>> problem_instance = vroom.Input()

>>> problem_instance.set_durations_matrix(
...     profile="car",
...     matrix_input=[[0, 2104, 197, 1299],
...                   [2103, 0, 2255, 3152],
...                   [197, 2256, 0, 1102],
...                   [1299, 3153, 1102, 0]],
... )

>>> problem_instance.add_vehicle([vroom.Vehicle(47, start=0, end=0),
...                               vroom.Vehicle(48, start=2, end=2)])

>>> problem_instance.add_job([vroom.Job(1414, location=0),
...                           vroom.Job(1515, location=1),
...                           vroom.Job(1616, location=2),
...                           vroom.Job(1717, location=3)])

>>> solution = problem_instance.solve(exploration_level=5, nb_threads=4)

>>> solution.summary.cost
6411

>>> solution.routes.columns
Index(['vehicle_id', 'type', 'arrival', 'duration', 'setup', 'service',
       'waiting_time', 'location_index', 'id', 'description'],
      dtype='object')

>>> groups = solution.routes[solution.routes.type == "job"].groupby("vehicle_id")["id"].apply(lambda x: set(x.dropna().astype(int)))
>>> groups[47] == {1414, 1515} and groups[48] == {1616, 1717}
True

Usage with a routing engine

>>> import vroom

>>> problem_instance = vroom.Input(
...     servers={"auto": "valhalla1.openstreetmap.de:443"},
...     router=vroom._vroom.ROUTER.VALHALLA
... )

>>> problem_instance.add_vehicle(vroom.Vehicle(1, start=(2.44, 48.81), profile="auto"))

>>> problem_instance.add_job([
...     vroom.Job(1, location=(2.44, 48.81)),
...     vroom.Job(2, location=(2.46, 48.7)),
...     vroom.Job(3, location=(2.42, 48.6)),
... ])

>>> sol = problem_instance.solve(exploration_level=5, nb_threads=4)
>>> 3800 < sol.summary.duration < 4200
True

Installation

Pyvroom currently makes binaries for on macOS and Linux. There is also a Windows build that can be used, but it is somewhat experimental.

Installation of the pre-compiled releases should be as simple as:

pip install pyvroom

The current minimal requirements are as follows:

• Python at least version 3.10.

• Intel MacOS (or Rosetta2) at least version 15.0.

• Apple Silicon MacOS at least version 15.0.

• Windows on AMD64.

• Linux on x86_64 and Aarch64 given glibc at least version 2.34.

Outside this it might be possible to build your own binaries.

Building from source

Building the source distributions requires:

• Download the Pyvroom repository on you local machine:

  git clone --recurse-submodules https://github.com/VROOM-Project/pyvroom

• Install the Python dependencies:

  pip install -r pyvroom/build-requirements.txt

• Install asio headers, and openssl and crypto libraries and headers. For mac, this would be:

  brew install openssl@1.1
  brew install asio

  For RHEL:

  yum module enable mariadb-devel:10.3
  yum install -y openssl-devel asio

  For Musllinux:

  apk add asio-dev
  apk add openssl-dev

• The installation can then be done with:

  pip install pyvroom/

Alternatively it is also possible to install the package from source using Conan. This is also likely the only option if installing on Windows.

To install using Conan, do the following:

cd pyvroom/
conan install --build=openssl --install-folder conan_build .

Documentation

The code is currently only documented with Pydoc. This means that the best way to learn Pyvroom for now is to either look at the source code or use dir() and help() to navigate the interface.

It is also useful to take a look at the VROOM API documentation. The interface there is mostly the same.