A self-organizing model for decentralized virtual environments in agent-based simulation systems

For simulations to be meaningful, it is necessary to implement realistic models for both virtual agents and environment. A lot of attention has been given to the definition of accurate models for agents. Unfortunately not much has been done for the definition of virtual environments that mimic the complexity of real-world environments. The reason is twofold: 1) The construction of realistic virtual environments (also called open environments) is not a trivial task [3]. Such environments are inaccessible, non-deterministic, dynamic and continuous. 2) Realistic simulations involve the execution of a large-number of sensor-based perception agents in an open environment. Unfortunately, limited computational resources make this goal untenable on a single machine. A few MABS have proposed models for open virtual environments. Most of these models represent the environment as a single massive component that is managed by one control unit. Other models decompose the environment into regions that are also managed by a single control unit [4]. In both cases, centralized control creates a bottleneck and limits the scalability of the simulation. On the other hand, a very limited number of MABS have proposed a partitioned structure of the environment with control units managing specific spatial areas [2]. Unfortunately, these systems do not leverage several of the benefits enabled by decentralized control. In this paper we propose a model for the execution of large-scale MABS with open environments on a single host. In our approach, agents execute their behaviors and are not subjected to any resource management constraints (e.g., aggregation). The open environment has a decentralized structure that is supported by an underlying self-organizing system. During the execution of the simulation, virtual agents