面向大规模地理空间仿真的并行多智能体模型  

Parallel Multi-agent Model for Large-scale Geospatial Simulation

作　　者：潘理虎[1] 张乐 闫慧敏[2] 苏雅 PAN Li-hu;ZHANG Le;YAN Hui-min;SU Ya(College of Computer Science and Technology,Taiyuan University of Science and Technology,Taiyuan Shanxi 030024,China;Institute of Geographical Sciences and Natural Resources Research,CAS,Beijing 100101,China)

机构地区：[1]太原科技大学计算机科学与技术学院,山西太原030024 [2]中国科学院地理科学与资源研究所,北京100101

出　　处：《计算机仿真》2023年第10期458-462,共5页Computer Simulation

基　　金：中国科学院战略性先导科技专项(A类)子课题(XDA2310 0202);山西省研究生优秀创新项目(2020SY439)。

摘　　要：针对大规模复杂多智能体仿真模型计算性能的问题,构建了一种基于分解-映射智能体任务调度方法的仿真模型。以草原生态政策驱动多智能体模型为例,分别进行1、50、100、500次任务的并行实验,预测不同任务数量下气候因素、出入栏率、综合效益等影响因子的演变过程及模拟消耗时间规律。实验结果证明该并行计算方法可以与多智能体模型完全兼容,极大的提高了大规模地理空间模型的仿真效率,对大规模复杂系统的研究具有重要的参考意义。Aiming at the problem of computational performance of large-scale complex multi-agent simulation model,a simulation model based on Decomposition-Mapping agent tasks was constructed in this paper.Taking the grassland ecological policy-driven multi-agent model as an example,the parallel experiments of 1,50,100 and 500 tasks were carried out respectively to predict the evolution process of climate factors,entry and exit rate,comprehensive benefits,and other influencing factors under different tasks,and simulate the time consumption pattern.Experimental results show that the parallel computing method can be fully compatible with multi-agent model,which greatly improves the simulation efficiency of large-scale geospatial model,and has important reference significance for largescale complex system research.

关 键 词：多智能体建模 复杂系统仿真 并行计算 草原生态 高性能计算 

分 类 号：N945.13[自然科学总论—系统科学]