Congestion prediction for smart sustainable cities using IoT and machine learning approaches

Congestion on road networks has a negative impact on sustainability in many cities through the exacerbation of air pollution. Smart congestion management allows road users to avoid congested areas, decreasing pollutant concentration. Accurately predicting congestion propagation is difficult however, due to the dynamic non-linear behavior of traffic flow. With the rise of Internet of Things devices, there are now data sets available that can be used to provide smart, sustainable transport solutions within cities. In this work, we introduce long short-term memory networks for the prediction of congestion propagation across a road network. Based on vehicle speed data from traffic sensors at two sites, our model predicts the propagation of congestion across a 5-min period within a busy town. Analysis of both univariate and multivariate predictive models show an accuracy of 84-95% depending on the road layout. This accuracy shows that long short-term memory networks are suitable for predicting congestion propagation on road networks and may form a key component of future traffic modelling approaches for smart and sustainable cities around the world.

Automated summary

Congestion on road networks worsens air pollution, but smart congestion management can help reduce pollutant concentration. Long short-term memory networks are effective in predicting congestion propagation and may play a key role in future traffic modelling for smart and sustainable cities worldwide.