The traffic state of a road network can be described by a so called macroscopic fundamental diagram (MFD). The average production is related to the accumulation of a specific road network in this diagram. Studies have shown that the MFD could be used in an evaluation based method of a certain road network. A promising application field of the MFD is to introduce the MFD in the steering mechanism of a traffic controller. Due to increasing congestion on urban roads, extra road capacity might be needed. However, for economic reasons better usage of the current road capacity should be performed. Therewith, the intersection density in road networks has increased nowadays. One of the consequences is that the way of controlling a certain intersection influences the traffic state at another intersection. Therefore, traffic controllers should be connected when controlling traffic at individual intersections. In this thesis a subnetwork flow controller has been designed. A road network which is controlled in a hierarchical setting by a main controller can be split up in several subnetworks. By controlling the perimeter flows between the subnetworks, the traffic state of each subnetwork can be controlled. In this thesis, only traffic signals at intersections have been taken into account as the control units. The designed subnetwork flow controller had to contribute to three main objectives: 1. Maintaining a constant shaped MFD, 2. Optimizing internal flows, 3. Provide desirable perimeter flows. The subnetwork flow controller algorithm has been based upon a back pressure algorithm which belongs to the coordinated traffic responsive control strategies in existing traffic controllers. A back pressure algorithm has been chosen due to the property of balancing queues which should result in homogeneity of traffic conditions within a subnetwork. The back pressure algorithm determines pressures for every individual intersection and every traffic phase consisting of several traffic streams. For every traffic stream the downstream queue length is subtracted from the upstream queue length and multiplied with the turn ratio at which traffic can go through the intersection at that specific traffic stream. The pressure of a phase is calculated by adding up individual pressures of traffic streams which are part of that specific phase. Due to the property of balancing queues by the back pressure algorithm and the assumption that homogeneity in traffic conditions might improve internal flows, some adjustments had to be performed only in order to provide desirable perimeter flows. A maximum deviation factor has been set up which allows a certain deviation of the actual perimeter flows with respect to the desirable perimeter flows which have been set up by the main controller. When the deviation exceeds a certain value, traffic streams have to be blocked when the actual perimeter flow is too high or have to get right-of-way when the actual perimeter flow is too low. By reducing the available phases from which the subnetwork flow controller can choose, the perimeter flows can be controlled. In order to evaluate the performance of the designed subnetwork flow controller, simulations have been performed in VISSIM where the subnetwork flow controller (written in Matlab) has been applied. Simulations with an applied vehicle-actuated controller and basic back pressure controller have been performed first in order to derive a most desirable size of the subnetwork layout and get reference results for evaluating the performance of the subnetwork flow controller. Simulations have been performed with subnetworks consisting of four, eight and sixteen intersections and different applied demand patterns. It turned out that a subnetwork consisting of sixteen intersections controlled by a vehicle-actuated controller or back pressure controller provides a MFD with the lowest scatter size, determined by the standard deviation of the absolute scatter deviation with respect to the constructed running median of the MFD. This low scatter resulted also in a constant shape of the MFD independent of the applied demand pattern. Therewith, with increasing size of the subnetwork the back pressure controller was able to control the traffic in the subnetwork better when evaluating total delay and internal production. This is caused by the result that gridlocks can be postponed by the back pressure controller. When applying the designed subnetwork flow controller on a subnetwork consisting of sixteen intersections, different maximum deviation factors have been applied. It turned out that there was no significant difference in performance on all three objectives between the applied deviation factors. Moreover, it turned out that the subnetwork flow controller was able to control perimeter flows at intersections with two adjacent intersections better as three adjacent intersections. Therefore, some extra simulations have been performed with an additional value of the maximum deviation factor and different desired perimeter flows for perimeter flows at intersections with two or three adjacent intersections. The subnetwork flow controller designed in this thesis has been proven to work properly according to the simulation results. When the subnetwork flow controller provides the desired perimeter flows (under certain circumstances), a constant shaped MFD can be maintained. However, another result is an increase in delay and thus less optimal internal flows compared to the performance of a vehicle-actuated controller. Therewith, a non constant shaped of the MFD was the result of some simulations caused by the influence of setting up the values for the desired perimeter flows and maximum deviation factor. No optimal values for the maximum deviation factor and restrictions on setting up the desired perimeter flows have been found in the algorithm of the subnetwork flow controller. It is therefore recommended to perform extra simulations in order to derive these aspects before the subnetwork flow controller is suitable in a hierarchical control structure. It is also recommended to perform extra simulations with more different kind of demand patterns and different control time intervals. Furthermore, future research is recommended on clearance times, the measuring way of queue lengths, applying more heterogeneous subnetworks, applying different kind of dynamic traffic management (DTM) measures and evaluating method of scatter size.