Section: New Results
Traffic prediction for fluid traffic models
In this work, the linear switched model for traffic state estimation/prediction based on the so-called Lighthill-Whitham-Richards (LWR) traffic flow model was developed. This model uses real traffic data as inputs and the dynamics of the model switching is given by a nondeterministic finite automata. The model performance was validated with respect to the performance of the original LWR traffic flow model as well as with respect to the performance of the microscopic traffic models (with use of the traffic simulator Aimsun 6). Various traffic state estimation/observation methods for the model have been tested. The so-called particle filtering seems to be a very efficient method for the problem, as it combines high observation/prediction accuracy with high computational speed. Fast and accurate traffic state estimation/prediction is important for use of suitable controls, such as dynamic tolls, on which we elaborated in  ,  .