1. An overview of AequilibraE

1.1. Sub-modules

AequilibraE is organized in submodules that are often derived from the traditional 4-step model. However, other modules have already been added or are expected to be added in the future. The current modules are:

Contributions can be made to the existing modules or in the form of new modules.

1.2. AequilibraE Project

The AequilibraE Project, as a consistent model file, comes from the vision of having a complete model system that would have capabilities to support the vast majority of analysis usually performed with traditional transport models.

In a nutshell, the AequilibraE project file is designed to behave like many of the commercial platforms available in the market, where a single file hosts the majority of the data. Differently then these platforms, however, AequilibraE is designed on top of an open data format, SQLite.

This concept is still under heavy development, but it is already possible to use its structure to download full, routable networks directly from Open Street Maps .

More detailed is provided in - The AequilibraE project.

1.3. Global parameters

As more features are added to AequilibraE, a large number of parameters start to be required, so the parameters module has also been growing in importance within the software.

There are currently 4 main sessions with the parameters file: Assignment, Distribution, Network and System.

The parameters for assignment and distribution control only convergence criteria, while the System section controls things like the number of CPU cores used by the software, default directories and Spatialite location for Windows systems. The Network section, however, contains parameters that control the creation of networks and the import from Open Street Maps.

The parameters file is built using the YAML format, and detailed instructions on how to configure it are provided on Parameter File.

1.4. Trip distribution

The trip distribution module is the second oldest piece of code in AequilibraE, and includes only code for calibration and application of Synthetic gravity models and Iterative Proportional Fitting. Not much documentation has been written for this module, but a some examples are available on the usage examples page Trip distribution.

1.5. Path computation

The path computation module contains some of the oldest code in AequilibraE, some of which preceed the existance of AequilibraE as a proper Python package.

The package is built around a shortest path algorithm ported from SciPy and adapted to support proper multi-threading, network loading and multi-field skimming.

As of now, this module has three main capabilities:

  • Traffic assignment

  • Path computation

  • Skimming

It is worth noting that turn-penalties and turn-prohibitions are currently not supported in AequilibraE, and that there is no built-in support for multiple concurrent shortest paths computation.

A wealth of usage examples are available in the examples page under Paths module.

1.6. Transit

For now the only transit-related capability of AequilibraE is to import GTFS into SQLite/Spatialite. The results of this import is NOT integrated with the AequilibraE project.

Usage examples can be found on GTFS import.

1.7. Matrix

The matrix submodule has two main components. Datasets and Matrices

Their existence is required for performance purposes and to support consistency across other modules. It also make it a lot faster to develop new features. Compatibility with de-facto open standards is also pursued as a major requirement.

They are both memory mapped structures, which allows for some nice features, but they still consume all memory necessary to handle them in full. In the future we will look into dropping that requirement, but the software work is substantial.

1.7.1. AequilibraE Matrix

If one looks into how all commercial software handle matrices, it would be clear that there are definitely two different schools of thought. The first one, is where matrices are those where matrices are simple binary blobs that require the existence of a model to give it context.

The second one is a more comprehensive take, where a matrix file not only contains its indices and metadata, but one that can also store multiple matrices. This capability, available in the openmatrix format, is one that was reproduced in the AequilibraE matrix and API.

Because the Open Matrix Format has established itself as the de-factor standard for matrix exchange in the industry, AequilibraE aims to allow users to never touch the AEM data format if they so decide. However, all underlying computation will be done using our custom format, so the importing/exporting being done under the hood will mean additional (although small) overhead.

For programatic applications wher eperformance is critical, we recommend using the AEM format whenever possible.

1.7.2. AequilibraE Data

AequilibraE datasets are data structures based on NumPy record arrays (arrays with named columns). Its role in the software is to hold columnar data for all procedures that may use it, such as O/D/P/A vectors used in trip distribution and link loads from traffic assignment.

AequilibraE data currently supports export to csv and sqlite. Extending it to other binary files such as HDF5 or Arrow are being considered for future development. If you require them, please file an issue on GitHub.