In real-time crowd simulations, the complexity of indexing the virtual entities should be kept low at all times in order to maximize the simulation performance. In this paper, a novel indexing technique is proposed. The proposed system is based on the navigation graph that is already commonly used for path finding in the current virtual crowd simulations. By this way, the simulation does not need a separate data structure for indexing the virtual entities, and at the same time it performs faster than the traditional grid based simulations that have narrow and long corridors (such as urban environments). In addition, while indexing static obstacles, the proposed system compresses and embeds the static obstacle information into the graph structure thus requires less memory than the traditional systems. The performance is evaluated for both the proposed and the traditional grid based systems on a virtual city environment. As a result, it has been shown that the proposed system on average works 8% faster than the traditional grid based system and requires significantly less memory while indexing static obstacles.