The pedestrian evacuation problem is highly stochastic due to the extreme uncertainty associ- ated with the failures caused by natural and man-made disasters. Thus, adopting a predetermined evacuation plan may be inadequate. On the other hand, it is difficult to design evacuation plans to address every possible scenario of failure. The availability of real-time information about the pedestrian network becomes vital for timely deployment of effective and case-specific evacuation strategies. In addition to the structural failures, operational failures are at hand due to the un- predictability of human behavior and decision making ability under stress rendering certain links within the network unusable. Therefore, in addition to maximizing the outflow of the network, preserving network controllability in real-time under the synthesized evacuation plans becomes cru- cial. The main objective of this proposal is to synthesize evacuation subnetworks in real-time that maximize outflow and maintain controllability of the pedestrian network based on multi-layered estimation schema.