API reference#

The API reference provides an overview of all public objects, functions and methods implemented in TASI. All classes and function exposed in tasi.* namespace plus those listed in the reference are public.

Datasets#

Traffic data is represented with different Dataset variants that are available via tasi.* or tasi.dataset.*.

Trajectory data#

While a Dataset allows to represent traffic data, TASI contains additional models to represent trajectory data. For that purpose, an hierarchical view is used and combined with a the tabular representation used by pandas.

Numerical representation#

When working with trajectory data in TASI, you will encounter the following three representation formats:

The TrajectoryDataset is a specific variant of a Dataset with trajectory specific functions. In fact, it is a collection of Trajectory objects. A Trajectory is, as the name indicates, the representation of a traffic participant’s trajectory. A trajectory is also a collection as the Dataset, but instead of trajectories it contains Pose objects. That is, a Pose is the traffic participant’s representation for a specific point in time. This distinction enables to provide additional functionality that is specific to the different representation variants.

Geospatial representation#

The models above are all based on pandas and positional information is typically encoded in two columns using UTM coordinates. Since trajectory data always contains positional information, TASI also provides models that allows to represent this information as GeoObjects. To achieve this, the three following models use geopandas instead of pandas:

This geopandas based representation of trajectory data allows to use the wonderful library shapely for computational geometry.

Pydantic representation#

To exchange data between TASI and other tools, TASI comes shipped with a representation layer that is build on pydantic and that enables conversion between different data formats. The models are available via tasi.io.public package.

Object-Relational Mapper models#

The representation level based on pydantic is levereged to convert TASI entities to other data formats. Moreover, since SQLModel is used for defining the attributes above, all the entities can be saved into a database using both SQLModel and SQLAlchemy. The models are available via tasi.io.orm package.

DLR specific models#

There are some specific classes that may help to work with the DLR datasets that are available via tasi.dlr.*.

Visualization#

TASI provides various tools that may help with visualization traffic data.