2. Getting Started¶
This page describes how to get started with AequilibraE.
Although AequilibraE is under intense development, we try to avoid making breaking changes to the API. In any case, you should check for new features and possible API changes often.
Install Python 3.6, 3.7 or 3.8. We recommend Python 3.7 as of January 2021.
pip install aequilibrae
Aequilibrae relies on a series of compiled libraries, such as NumPy and Scipy. If you are working on Windows and have trouble installing any of the requirements, you can look at Christoph Gohlke’s wonderful repository of compiled Python packages for windows.
22.214.171.124. OMX support¶
AequilibraE also supports OMX starting on version 0.5.3, but that comes with a few extra dependencies. Installing openmatrix solves all those dependencies:
pip install openmatrix
Although the presence of Spatialite is rather obiquitous in the GIS ecosystem, it has to be installed separately from Python or AequilibraE.
This blog post has a more comprehensive explanation of what is the setup you need to get Spatialite working, but below there is something you can start with.
Spatialite does not have great support on Python for Windows. For this reason, it is necessary to download Spatialite for Windows and inform AequilibraE of its location.
After unpacking the zip file into its own folder (say D:/spatialite), one can start their Python session by creating a temporary environment variable with said location, as follows:
import os from aequilibrae.utils.create_example import create_example os.environ['PATH'] = 'D:/spatialite' + ';' + os.environ['PATH'] project = create_example(fldr, 'nauru')
For a permanent recording of the Spatialite location on your system, please refer to the blog post referenced above or Windows-specific documentation.
126.96.36.199.2. Ubuntu Linux¶
On Ubuntu it is possible to install Spatialite by simply using apt-get
sudo apt-get install libsqlite3-mod-spatialite sudo apt-get install -y libspatialite-dev
2.2. Hardware requirements¶
AequilibraE’s requirements depend heavily of the size of the model you are using for computation. The most important things to keep an eye on are:
Number of zones on your model (size of the matrices you are dealing with)
Number of matrices (classes you are dealing with)
Number of links and nodes on your network (far less likely to create trouble)
Substantial testing has been done with large real-world models (up to 8,000 zones) and memory requirements did not exceed the traditional 32Gb found in most modelling computers these days. In most cases 16Gb of RAM is enough even for large models (3,000+ zones). Parallelization is fully implemented for path computation, and can make use of as many CPUs as there are available in the system when doing traffic assignment.