The increase of mobility of the past decades has led to substantial congestion on the freeways. Traffic jams emerge both on a daily basis at the same location, as well as during accidents when a part of the freeways is temporarily blocked. In those cases, traffic management centers intervene into traffic in order to reduce or even dissolve congestion. This is called Dynamic Traffic Management (DTM). Common DTM measures are rerouting traffic, ramp metering at the on-ramps of freeways and opening peak-hour lanes. DTM is performed in two steps. First, the current traffic state is estimated by fusing sensor data, usually from dual-inductive loops or from in-car sensors. The traffic state describes where the traffic is located in the network and how fast it travels. Second, based on the current traffic state, a controller determines appropriate actions for each DTM measure in order to improve the traffic performance. In current practice, DTM controls traffic as a whole, not differentiating between the different vehicle classes. However, vehicles can be classified according to length, maximum speed or value of time. The vehicle classes therefore have different effects on the network performance. For example, short vehicles can travel with a shorter time headway than longer ones. Consequently, more short vehicles can pass any given location than long one. The capacity for shorter vehicles is therefore larger than for long ones. In this thesis, Dynamic Traffic Management is generalized to take the properties of different vehicle classes into account. The effects of the vehicle classes on the traffic flow and the network performance are analyzed based on the macroscopic multi-class traffic flow model Fastlane. Furthermore, existing DTM measures are generalized in order to control traffic vehicle-class specifically. A multi-class ramp meter is developed that is able to meter each vehicle class individually. Prioritizing short vehicles increases the network throughput; conversely, prioritizing valuable vehicles decreases the total cost. Multi-class route guidance enables the routing of a vehicle class around a congested area. At bottleneck locations, a class-specific lane makes it possible to keep a specified vehicle class in free-flow. In order to apply DTM in real-time, two existing traffic state estimators are analyzed and reformulated so that they now estimate the traffic state of realistically-size freeways within a few seconds. The Adaptive Smoothing Method is reformulated to use the Fast Fourier Transform. The Extended Kalman Filter technique is localized so that measurements are used to correct the system state only in the vicinity of the measurement instead of correcting the state of the whole network. Furthermore, a tool is developed that extracts the position and speed of shock waves from spatiotemporal traffic data, which are used to calibrate traffic state estimators or traffic flow models. The developed components of estimation and control are then combined in a case study to optimize class-specific control advices for the Dutch freeway A15 near the harbor of Rotterdam. The case study shows that multi-class DTM improves the network performance compared to conventional, mixed-class DTM. Currently, those control advices are calculated continuously and are published at a website in real-time. In addition, the current traffic state, the future traffic for the next hour without control, and the future traffic with multi-class control are published.