Prediction of traffic flow for real-time control

Abstract

The prediction of traffic flow on a network and the relationship of these flows to the traffic control signal settings are major factors in the development of an adaptive real-time signal system. PASSER IV is an arterial system optimization package which generates optimal signal timing plans using traffic flows as inputs. To use PASSER IV in a real-time signal control system, it is essential to accurately predict the traffic flow conditions for the next control period. The objective of this study was to propose a prediction algorithm and test the same with existing models to select the most appropriate one for field implementation. The prediction model developed was tested along with exponential smoothing and historic average predictors at the intersection of Beltline and Plano in Richardson, Texas. The test results concluded that the use of historic data was a good predictor where traffic patterns remain steady. The proposed algorithm, which was an adaptation on historic average, performs as good and in some cases better, as it is able to adjust to changing traffic patterns by using on-line data. Exponential smoothing does not perform well where there are sudden highs and lows as often observed in arterial traffic patterns in Richardson. It tends to underestimate when there is a sudden increase and overestimate in case of a sudden decrease in traffic volumes. If the double exponential forecasting method can be made adaptive, it would prove useful where historic data is non-existent.