This thesis presents a simulation model of the evacuation process where the evacuee uses the architectural cues to search his route to the safety in a space design of a complex public underground environment. With the booming urbanization, new mega cities have to develop the underground space network. Every node in this network always consists of subway station and connections to the surrounding commercial underground spaces, which is public in function and complex in spatial organization. To support the development of such a complex public underground space, the performance-based evaluation method using computer-based evacuation simulation replaced the traditional prescriptive regulation-based method. However, with a vague understanding on how space influences the evacuee’s route searching behavior, the architects still haven’t a proper evaluation tool for their space design. Obviously, the space design is the fundamental part of the whole design process and the main task of the architect’s work. Thus, the research in this thesis aims to provide a clear understanding on such a space-design-based route searching process and to present a simulation model for the architect to evaluate his space design of complex public underground environment. According to the literatures, escape route searching is a kind of way-finding process. The evacuee searches his route to the safety in a stepwise decision process. In every step, he scans the local architectural cues in his view, chooses the cue with the highest possibility leading to the safety by his estimation, and moves to the direction hinted by the chosen cue. With the above theory, several methods were used to build a simulation model. First, questionnaires were used to survey from which architectural element the evacuees can perceive the architectural cues in this research context. Second, a CAVE platform was built to observe the virtual evacuation behavior to deduce the knowledge about the cognition rules that are applicable for local architectural cues. Third, the simulation model framework was composed for the route searching process. Fourth, the CAVE-based Conjoint Analysis approach was used to build the preference function to transit the qualitative conclusions on the local architectural cues to the quantitative variables driving the simulation model. Finally, with all the conclusions of these researches, the computer program SpaceSensor was built. In summary, this research has resulted in a new simulation model for architectural cue driven evacuation behavior. It contributes to the architect’s evaluation on the space design of complex public underground environment through the simulation of the route searching process and an architect-oriented safety criterion LEEI based on this simulation. Moreover, with all the details in the route searching process explained quantitatively, it also contributes to the architect’s understanding on how his space design influences evacuation.