Traffic Flow Modelling with Point Processes

The main objective of this article is to predict the traffic flow for the purpose of urban traffic management. Long term incidents, such as floods, road maintenance or major traffic incidents can cause significant disruptions on the traffic flows in large metropolitan areas such as Sydney. There is a need to comprehend how traffic will be impacted in a city during such an event. One main reason for this is to manage the flow of this traffic so that it does not obstruct with emergency protocols. Such modelling requires the capturing of fast-evolving and ever changing conditions, taking into account road closures, Twitter or Facebook information as well as other circumstances. One important characteristic in traffic flow modelling is the ability to describe traffic burstiness, i.e. vehicles tend to cluster together, thus creating traffic congestion. This paper provides some promising results in regards to traffic flow modelling using point processes framework. We are able to simulate traffic flow that mimics real life conditions, and also perform efficient inference thereof.