In this thesis a new ATCS prototype has been developed and evaluated that is directly based on two previous research projects. The first of these two projects (ALMASRI, 2006) dealt with model-based offset optimization for signalized intersections in an urban network under the regime of traffic signal control. The project used GA in combination with the CTM. The second project (WANG, 2008) dealt with the estimation of OD matrices and the influence of redundant estimation constraints, i.e. redundant traffic counts, on the quality of the estimation result. Several rules had been developed in this project to eliminate redundant constraints. Furthermore, the process of OD matrix estimation had been incorporated into an iterative procedure of traffic assignment and matrix estimation until convergence of the estimated matrix on a stable solution is achieved. Before discussing all modules of the new ATCS prototype in detail, an overview of the conceptual design of the prototype has been given. It applies a centralized concept, i.e. signal timings of all signalized intersections are optimized on a central computer. The optimization interval of the strategy is 15 minutes. At the end of each interval, the signal controllers at each intersection send current counts from their detectors to the ATCS central computer where they are used as constraints for the estimation of the current traffic demand. Based on this estimated demand, the ATCS adjusts a common cycle length, phase durations and offsets of all intersections. After termination of this optimization procedure, the new signal timings are sent back to the controllers at the beginning of the next time interval. There, the new signal timings are implemented immediately. One of the basic elements of the ATCS is the traffic flow model which is used to estimate the impact of different offset combinations in terms of total delay. The CTM has been used for this purpose. The original model equations have been presented thoroughly, followed by some extensions developed explicitly in this thesis to enable better modeling of urban intersections of arbitrary geometry. Modeling of permitted left turns has also been included into the CTM. Besides the model-based estimation of delays, an additional approach has been developed and tested to estimate travel times on different routes in the network. The model has been carefully validated. It could be shown that the produced results are in good accordance with the results of the microsimulator AIMSUN, and that the CTM can be used to deliver reliable, even though not perfect estimates of delay and travel times in an urban network. Since the ATCS prototype needs an estimate of the upcoming traffic demand of the next optimization interval in order to be able to perform an optimization of signal timings, the first two modules of the ATCS prototype cover traffic demand estimation. The first module performs a forecasting of detector counts. An approach by FÖRSTER (2008) has been implemented and tested for this purpose. The approach relies on space-time-patterns of detector counts of the four previous time intervals. The current pattern, which represents the traffic demand that has been observed during the last hour, is searched for in several reference patterns (in this thesis, only one such reference pattern has been used). The best matching sub-pattern within the reference pattern is identified and the values of the following time interval in the reference pattern can then be used to calculate estimates of the forecasted detector counts. In this thesis it could be shown that from time to time systematic over- or underestimation of forecasted counts occurs at some time intervals. The reason for these systematic errors has been investigated and the method has been slightly adapted. The effect could be reduced, but not eliminated entirely, i.e. some imprecision of the forecasting module remained.