Grey models for short-term queue length predictions for adaptive traffic signal control

Abstract

Traffic congestion at a signalized intersection greatly reduces the travel time reliability in urban areas. Adaptive signal control system (ASCS) is the most advanced traffic signal technology that regulates the signal phasing and timings considering the traffic patterns in real-time to reduce traffic congestion. Real-time prediction of traffic queue length can be used to adjust the signal phasing and timings for different traffic movements at a signalized intersection with ASCS. The accuracy of the queue length prediction model varies based on many factors, such as the stochastic nature of the vehicle arrival rates at an intersection, time of the day, weather, and driver characteristics. In addition, accurate queue length prediction for multilane, undersaturated and saturated traffic scenarios at signalized intersections is challenging. Thus, the objective of this study is to develop short-term queue length prediction models for signalized intersections that can be leveraged by adaptive traffic signal control systems using six variations of Grey systems: (i) the first-order single variable Grey model (GM(1,1)); (ii) GM(1,1) with Fourier error corrections (EGM); (iii) the Grey Verhulst model (GVM), (iv) GVM with Fourier error corrections (EGVM), (v) the Grey model with cosine term (GM(1,1|cos(ωt) or GMC), and (vi) GMC with Fourier error corrections (EGMC). The efficacy of the Grey models is that they facilitate fast processing; as these models do not require a large amount of data; as would be needed in artificial intelligence models; and they can adapt to stochastic changes, unlike statistical models. We have conducted a case study using queue length data from five intersections with adaptive traffic signal control on a calibrated roadway network in Lexington, South Carolina. Grey models were compared with linear, nonlinear time series models, Adaboost (ADA), Bagging (BAG), Gradient Boost (GB), Neural Network (NN), Random Forest (RF), Support Vector Regression (SVR), and Long Short Term Memory (LSTM) models. Based on our analyses, we found that GMC can produce better or competing performance compared to more complex (i.e., LSTM and NN models) in predicting average and maximum queue lengths in terms of root mean squared and mean absolute errors. GMC model needs only 4 data values and can be used in real-time as the computational time to generate a prediction in less than 0.001 s.