/FleetPy

Simulation framework to model and control the tasks of vehicle fleets (routing, user-assignment, charging, ...)

Primary LanguagePythonMIT LicenseMIT

FleetPy

Simulation framework to model and control the tasks of vehicle fleets (routing, user-assignment, charging, ...)

Features

  • Agent-based simulations of dynamic vehicle routing problems

  • Time- and event-based user-operator interaction models

  • Simulation contains information flow, i.e. request – offer – accept/reject

  • Various request-acceptance models, e.g. ‘always accept’, ‘accept if hard constraints are satisfied’, ‘probabilistic sensitivity on offer parameters’, or even ‘probabilistic based on mode-choice model’ that considers also other mode alternatives

  • Several fleet control strategies to assign requests, pool and assign requests, reposition vehicles, re-charge electric vehicles

  • Different underlying routing algorithms, choice can be made based on trade-off between memory requirement and computation time

  • Modular approach enables combination of different request models, network representations, fleet control strategies

Dependency installation

Conda workflow for virtual environment

Prerequisite: anaconda installed; Example in this instruction: Anaconda version: 201903; conda version: 4.9.2

After installing Anaconda, we could open the Anaconda Prompt to execute the following codes. First, let's change the working directory.

cd <working_directory>

Control the channel priority

It is strongly recommended by GeoPandas to either install everything from the defaults channel, or everything from the conda-forge channel. Ending up with a mixture of packages from both channels for the dependencies of GeoPandas can lead to importing problems.

In this instruction, we choose the channel conda-forge.To achieve this, first, we could add the channel conda-forge by

conda config --env --add channels conda-forge

You should check all your channels by:

conda config --show channels

If conda-forge is not on top of your list, please add the channel once more to put it on top of the list.

To restrict the channel, use the following code:

conda config --env --set channel_priority strict

Which basically installs packages with same names strictly by channel priority, and as it is on top of the list, the higher channel is the conda-forge. To check more detailed information about the channel priority, input the following code:

conda config --describe channel_priority

Create the virtual environment

Create a new virtual environment in an Anaconda Prompt with a Python version 3.7.X.

conda create -n <new_env> python=3.7

Checkout to the created virtual environment

conda activate <new_env>

Check the default packages in your new and clean virtual environment 😄

conda list

Install packages direcly using Conda:

conda install --file requirements.txt

It works! Looks like every package is successfully installed! Now check the installed packages again 😉

conda list

Everything is set up! 👍 Now you could run your first simulation!

C++ Router

We recommend to use the C++ router unless your network is small enough to preprocess the complete travel time matrix. If you want to use the C++ router, you need to have a C++ compiler and Cython set up on your computer.

Moreover, you need to compile the module on your system. Please go to

cd FleetPy/src/routing/cpp_router

Next, you can install the C++ router:

python setup.py build_ext --inplace

Optimizer

  • Gurobi: Set gurobi channel on top of your channel list by twice calling
conda config --add channels http://conda.anaconda.org/gurobi

Install gurobi package by

conda install gurobi

Free academic licenses of Gurobi can be acquired. See https://www.gurobi.com/academia/academic-program-and-licenses/ for more details in installation instructions.

Data Preparation and Study Setup

For now, you can inspect the data structures and files in the examples provided in github:

  • FleetPy/data
  • FleetPy/studies

More detailed descriptions of the data structure, preprocessing steps, and result data will be provided in the next versions. Additionally, a GUI to set up scenarios (with choice of submodules and data) is planned for the future.

In general, you can save your data and study definitions in the mentioned directories. These are included in .gitignore.

Usage

Simulation

You can test the example scenarios provided in the github repository by calling

python3 run_examples.py

You should now have a directory containing several simulated scenarios. In each of the scenario directories, you will have a summary of the configuration, a simulation log file and several output files.

  • FleetPy/studies/example_study/results The output files will be described in more detail in future versions; a very brief description for now:
  • 1_user_stats.csv contains user data for every single request
  • 2_0_op-stats.csv contains vehicle data for every task of the fleet vehicles of operator 0

By default, a standard evaluation aggregating several results is performed after every simulation.

  • standard_eval.csv Additionally, you can create a few temporal evaluations by calling
python3 FleetPy/src/evaluation/temporal.py scenario_result_directory

When you want to run your own scenarios, please create a copy of this script file, denote it 'run_private_[XYZ].py' with any study indicator XYZ, and modify the part following

if __name__ == '__main__'

'run_private*.py' is included in .gitignore.

Watch Replay

If you want to see a visualization of an already simulated scenario, you can call

python3 replay_pyplot.py scenario_result_directory sim_seconds_per_real_second

The start time of the replay can be given as an additional optional input parameter.

Tested on

Windows 10 Pro x64 Chrome 79.0.3945 Python 3.7