Traffic Flow and Deep Neural Networks

Most inter- and intra-city transportations occur via vehicles. Managing the traffic caused by such vehicles is of great importance for several primary reasons. Proper traffic control saves plenty of time and resources, assures the safety of drivers and passengers, and last but not the least, could help mitigate the issue of air pollution.

While the modern trend towards producing electric vehicles (EVs) is rising, a total change of fleet to EVs will require a substantial amount of time. For these reasons, the traffic flow at E4 Southbound has been the subject of a collaboration between ENCCS, KTH, and Trafikverket.

As a case study, the traffic flow will be modeled on a specific part of the E4S using deep neural networks. The results will give us a hint at the most appropriate models for such modeling. In the second part of the project, the already available model Temporal Graphical Convolutional Networks (T-GCN), will be tested and ported to an HPC system for further analysis of a much larger dataset corresponding to the E4S more extensive structure.

This undertaking is a collaboration between Assoc. Prof. Xiaoliang from KTH, Mr. Christian Edfjall from Trafikverket, and Dr. Hossein Ehteshami from ENCCS.