Robust real-time traffic management methodologies are needed to respond to freeway congestion caused by unexpected incidents. A multistage stochastic mathematical model with recourse is introduced to compute and disseminate real-time traffic control actions, which account for system uncertainties such as demand variation and incident severity. The benefits include the provision of a priori robust traffic control actions that are ideal under all circumstances of system uncertainties across all considered time periods, with significant reduction in the online computational effort. As a result, more insights on the relationship between traffic control actions and network properties across time periods are obtained.