Traffic queue estimation

This paper deals with the urban traffic queue modelling and estimation. A general formulation of delays on a road section is used to link measured speeds at its entrance to the current queue state. This measurement model is applied to correct a posteriori the queue state distribution whose a priori evolution is ruled by a time varying Markov chain.The resulting predictor-corrector is tested with a microscopic simulation tool. Results in fixed plan control show that for long road sections (200 meters) the estimation error is unacceptable whereas for short road sections (50–100 meters) the average quadratic error is about 1 vehicle2. In real time control conditions, in spite of a significant increase of the estimation quadratic error, results show a delay benefit of 15% with respect to the fixed plan control.