Observer models for real-time traffic control

The control of traffic lights today is mostly done by program selection depending on the daytime or of the average traffic flow. Flow models are the basis for the offline optimization. These rough models cannot be applied for real-time traffic control to achieve time-critical goals. In this paper different observer models for noiseless detectors are described. The identification of the model parameters is done. Verification and validation are subject of future work. Robustness can be obtained by an adequate error model.