Cycle-Length Prediction in Actuated Traffic-Signal Control Using ARIMA Model

Abstract

In urban transportation systems, the traffic signal is the main component in controlling traffic congestion. Using actuated traffic control as one of the traffic-controlling systems can cause fewer delays for transportation users, specifically when it comes to an isolated intersection. Although actuated signal control has many benefits, the prediction of cycle length is cumbersome because it varies from time to time. The value of signal cycle length in actuated control depends on many parameters. In this research, the authors attempted to understand whether any dependence existed between the current value of the cycle length and its previous values. To capture the dependence among cycle length data, time series analysis was applied over the data, which were obtained from the simulated fully actuated signal. The behavior of the signal’s cycle length under different levels of demand was analyzed, and, based on sample autocorrelation functions (ACFs), a well-known family of time series called autoregressive integrated moving average (ARIMA) was chosen for model fitting and prediction. The results revealed that there is a statistically significant dependence between two consecutive cycle lengths, and this dependence becomes more pronounced as the demand increases. Further, to improve the fit and prediction accuracy of cycle length for signals with more than two critical phases, a linear regression component using skipping indicators has been added to the ARIMA model. Finally, simulation-based cycle length prediction using the proposed model performs reasonably well under different simulation scenarios, and it achieves a smaller mean squared prediction error (MSPE) as compared to more traditional averaging prediction models.