论文部分内容阅读
作为一种研究复杂系统的有效途径,基于Agent的建模仿真方法得到了广泛应用,但随着应用规模和复杂度的增加,仿真运行速度成为制约其应用的一个重要因素。并行Agent仿真通过将Agent模型分配到多个处理单元上同时运行来减少仿真运行时间,是提高仿真运行速度的一个直接手段。根据基于Agent的建模仿真的基本思想,分析了并行Agent仿真区别于传统的并行离散事件仿真的新特点,着重从时间同步协议、负载均衡、通信优化三方面阐述了并行Agent仿真的研究现状,并对基于通用目的图形处理器的并行Agent仿真进行了讨论。最后对并行Agent仿真未来的研究方向进行了展望。
As an effective way to study complex systems, Agent-based modeling and simulation methods have been widely used, but with the increase of application scale and complexity, the simulation speed has become an important factor restricting its application. Parallel Agent Simulation Distributes the Agent model to multiple processing units simultaneously to reduce the simulation run time, which is a direct measure to speed up simulation. According to the basic idea of agent-based modeling and simulation, this paper analyzes the new characteristics of parallel agent simulation which is different from the traditional simulation of parallel discrete event, emphasizes on the research status of parallel agent simulation from three aspects: time synchronization protocol, load balancing and communication optimization. The parallel agent simulation based on general purpose graphics processor is discussed. Finally, the future research direction of parallel agent simulation is prospected.