Short-term forecasting of uncertain traffic states

Short-term traffic flow forecasting is considered to be the core of intelligent transportation systems and the advanced technical support including traffic information service, traffic control, and guidance. Previous traffic information is on current or predicted traffic states, which are still deterministic and difficult to show with the uncertainty of the road network. Therefore, the analysis of reliability of short-term forecasting is proposed to deal with the uncertainty of traffic. The reliability of short-term forecasting is described as a range of forecast values which show the deviation between forecast value and the measure. A state space model is used in this paper and the Monte Carlo method is performed to deal with uncertainty, where a random sample of traffic flow within a probability distribution is generated until at a steady state. Results show that the proposed model can reduce the disturbances between the forecast and actual traffic. © 2009 ASCE.