Quick optimal lane assignment using artificial neural networks

Signalized intersections are major components for any transportation network as the traffic operation at these intersections will significantly affect the operation for the whole network. Dynamic lane assignment (DLA) represents an Intelligent Transport System (ITS) technique that can be used to enhance the traffic operations at signalized intersection by utilizing the space efficiently. In DLA strategy, the number of lanes assigned for each movement (left, through and right) depends mainly on the real time traffic demand for that movement. This study aims to develop an artificial neural network (ANN) model that can be used to predict the optimal lane assignment combinations at signalized intersections using turning movement volumes for all intersection approaches. Developing an ANN model will expedite the selection process for the optimal lane assignment since it does not require detail delay calculations for all possible lane combinations to identify the optimum lane configuration for a given traffic movement. The proposed ANN model had three hidden layers with 14 neurons and gave an average accuracy of 92% on the test dataset.