A Genetic Approach for Distributing Semantic Databases of Crowd Simulations

Last years have witnessed how crowd simulations have become an essential tool for many virtual environment applications. These applications require both rendering visually plausible images and managing the behavior of autonomous agents, and therefore they need a scalable design that allows them to simultaneously tackle these two requirements. One of the main problems in the design of a scalable crowd simulation consists of efficiently distributing among different computers the semantic database containing the virtual world. In this paper, we propose a genetic approach for distributing the semantic database of crowd simulations in such a way that the dependencies among the computers hosting the pieces of the database are minimized. The proposed approach avoids the saturation of these computers by ensuring that the size of the pieces assigned to each computer is properly balanced. The performance evaluation results show that the proposed approach significantly reduces the resulting overhead in regard to other local search methods, regardless of the movement pattern of the agents. Therefore, it allows an effective partition of the semantic database.