多Agent系统在空间直观仿真建模中的并行化

针对空间并行仿真技术中的同步问题,设计基于Agent感知区域的有限乐观同步算法。该算法继承乐观同步算法的积极性、保守同步算法的安全性,并将仿真中的Agent划分为安全Agent区域和非安全Agent区域,同时利用调度服务器和执行服务器分工协作的方式完成仿真计算。调度服务器根据不同Agent的安全半径筛选确定其是否属于安全区域,执行服务器负责执行Agent的行为。该算法实现仿真技术中积极且有限的乐观同步,降低并行仿真中网络的通信量,确保仿真程序执行过程中的因果约束关系,有效地缩短了仿真时间,在GAMA仿真平台实现并验证方法的有效性。

Parallel Computing of Multi-Agent System in Spatially-Explicit Modeling and Simulation

Aiming at the synchronization in the spatially-explicit parallel simulations, a limited optimistic synchronization algorithm based on Agent sensing region is proposed. The algorithm inherits the advantages of the optimistic synchronization algorithm and the security-guaranteed conservative synchronization algorithm. The algorithm categories all simulated agents into the security agent region and non-security agent region. The simulation is performed through the collaboration between the dispatch server and the execution servers where simulation tasks are scheduled and executed. The dispatch server determines whether an agent is in a security zone according to the security radius. The execution servers are responsible for performing the behavior of the agents. The algorithm achieves the better performance claimed in the limited optimistic synchronization, reduces the network traffic in the parallel simulation, ensures the causality constraint relationship in the process of simulations and decreases the simulation time cost. Experiments in GAMA platform demonstrate the effectiveness of the algorithm.