Estimating Road Traffic Congestion from Cell Dwell Time using Neural Network

Abstract

In this study, we investigated an alternative method to estimate the degree of road traffic congestion based on a new measurement metric called cell dwell time (CDT) using simple feedforward backpropagation neural network. CDT is the duration that a cellular phone is registered to a base station before handing off to another base station. As a vehicle with cellular phone traverses along the road, cell handoffs occur and the values of CDT vary. Our assumption is that the values of CDT relate to the degree of traffic congestion and that high CDTs indicate congested traffic. In this study, we measured series of CDTs while driving along arterial roads in Bangkok metropolitan area. Human judgment of traffic condition was recorded into one of the three levels indicating congestion degree -free flow, moderate, or highly congested. Neural network was then trained and tested using the collected data against human perception. The results showed promising performance of congestion estimation with accuracy of 79.43%, precision ranging from 73.53% to 85.19%, and mean square error of 0.44.