Congestion on the road is a widely recognized problem. To improve traffic conditions on the road, several methods have been developed over the years. Dynamic traffic management is one of them and has proven to be effective. It aims at making better use of the existing network capacity and at managing traffic flows. In order to further improve the effectiveness of the individual measures, research into coordination of individual traffic management measures (i.e. integrated network management) has recently been increased. Effectiveness can be improved since (i) the counteracting of measures against each other can be reduced and (ii) the strengths of the individual measures can be combined. To show the benefits of integrated network management in a real world situation, a well performed field operational test is needed. In the Netherlands this was a reason to launch the Field Operational Test Integrated Network Management Amsterdam, in Dutch the Praktijkproef Amsterdam. Part of the Praktijkproef Amsterdam was the development of a controller that can control bottleneck situations which occur in the neighbourhood of a junction of a freeway and an urban arterial. Bottleneck situations that can occur are: (i) spillback on the urban arterial causing blocking back on the urban arterial, (ii) spillback from the off-ramp towards the freeway causing congestion on the freeway, (iii) spillback from the on-ramp towards the urban arterial causing blocking back on the urban arterial. These three situations capture all possible bottleneck situations that can occur in this type of network. In the Praktijkproef Amsterdam a certain controller was developed which can only handle the third situation. Therefore, in this research a controller is developed that is able to control all three situations. Hence, the objective of this research is a controller that deals with bottleneck situations occurring in the neighbourhood of a junction of a freeway and an urban arterial; in order to reach its goals, the controller should distribute traffic over the available buffer space in the network, by using traffic lights at controlled intersections. In the first situation the bottleneck needs to be detected and controlled (detection is input for control), in the second and third situation the queue at the ramp needs to be controlled. The research consists of three phases: (i) a literature survey that studies the state-of-the-art related to the controller to be developed, (ii) the development and programming (in MATLAB) of different controller variants and (iii) the simulation (in VISSIM) of the variants. The literature survey showed that controllers that are capable of controlling the three bottleneck situations mentioned, do not exist at the moment. Several controller variants are developed to distribute traffic over the buffers in one of the bottleneck situations. Distribution is based on changing the signal settings of the traffic lights: an increase in green time results in a decrease in queue length, and vice versa. Signal settings are changed based on calculated desired flows for the buffers. For the first situation three detection variants are developed: detection based on (i) a crisp critical queue length value, (ii) differences in queue lengths between two time periods, (iii) fuzzy logic with queue lengths and flows as input values. Furthermore three controller variants are developed: controllers that distribute the surplus of traffic over (i) one up- or downstream buffer based on prespecified preferences of using up- or downstream buffers, (ii) one up- or three downstream buffers also based on these prespecified preferences of using up- or downstream buffers, (iii) one up- and one downstream buffer based on relative buffer space. For the second situation the queue at the off-ramp is managed by increasing the outflow at the ramp, based on a target outflow. Due to this increase spillback should be prevented. Two controller variants are developed: a controller which increases the outflow by distributing traffic over (i) the first downstream buffer, (ii) three downstream buffers based on turn fractions. For the third situation the queue at the on-ramp is managed by reducing the inflow into the ramp, based on a target inflow. This should lead to the prevention of spillback at the ramp. Three controller variants are developed: controllers that reduce the inflow by distribution traffic over the (i) first upstream buffer, (ii) three upstream buffers based on turn fractions, (iii) three upstream buffers based on relative queue lengths. The ST1Light (developed in the Praktijkproef Amsterdam) is the fourth variant in this bottleneck situation, the ST1Light calculates the amount of buffers needed to reach the target inflow. In the first situation congestion on the urban arterial is reduced by preventing spillback. A fuzzy logic approach using flow and queue length as inputs in order to detect bottleneck situations, shows best results in combination with the controllers: the smooth approach of the fuzzy detection results in more smooth control actions and therefore less variation in queue lengths at the bottleneck. The controller that prefers to use downstream buffers shows largest improvements in network performance (-3.3% in total travel time). The controller that uses both up- and downstream buffers has the strongest effect on reducing the queue at the bottleneck situation, but due to the upstream buffering total travel time increases (+1.7% in total travel time). In the second situation congestion at the freeway is prevented, by preventing spillback from the ramp towards the freeway. Both designed controllers prevent spillback towards the freeway, hence preventing congestion and the capacity drop at the freeway. There is no trade-off shown at the urban arterial. This results in large improvements in the overall network performance. If more downstream buffers are used, the traffic is flushed further into the network and it reaches the network boundaries faster, hence resulting in shorter travel times and larger improvements in overall network performance (-38.7% in total travel time). In the third situation spillback from the on-ramp towards the urban arterial is prevented by the developed controllers. If more buffer capacity is used, the network performance shows larger reductions due to the buffering of traffic. Therefore the controller which only uses the first upstream buffers, shows best results in network performance (-1.1% in total travel time). If no control is used, the network performance is better (-3.8% in total travel time), but in that case spillback from the on-ramp is not prevented. If no spillback occurs, the metering time of ramp metering installations will increase and traffic safety in the network will improve since the conflict area at the urban arterial of the on-ramp intersection is not occupied anymore. It can be concluded that if buffers downstream of the bottleneck can be used, the controllers show positive results regarding the network performance; and if buffers upstream are used, delays for traffic upstream of the bottleneck increase. The latter results in a decrease in network performance. First of all it is recommended to implement bottleneck detection on urban arterials, based on a fuzzy logic approach. Bottleneck detection is not a part of current intersection control systems, and can be added to those systems. In order to control the bottleneck situations, the developed controller that prefers to use downstream buffers can be coupled to a current control system. Secondly it is recommended to prevent spillback at the off-ramp by setting a target outflow. The controller that only uses the first downstream buffer can be combined with current active systems. Thirdly it is recommended to prevent spillback at the on-ramp by setting a target inflow. It is recommended to use the first and second upstream buffers to buffer traffic. Furthermore it is recommended to switch off the controller if spillback occurs to conflict areas of the intersections, since the latter leads to large increases in travel times at the urban arterial. Future research should focus on further tuning (e.g. fuzzy parameters) and testing (e.g. different traffic conditions, increased network size) of the control algorithms, and on combining the controller with current used systems in the field. Furthermore a supervisor could be created that can deal with multiple active bottleneck situations.