摘要
进化仿真实验是人工生命领域研究的主要途径,其成果可广泛应用于多主体演化计算、生化过程仿真、军事演练、群体行为仿真、复杂系统研究等,而其计算量往往很大。为比较全面地总结人工生命仿真领域的主要研究特点、方向和难点等问题,对现有进化仿真相关文献通过文献计量、摘要等系统化综述手段进行分类、比较和概括,发现:为解决进化仿真的大计算量问题,需要针对其紧密时空耦合的计算特点,借鉴数据库技术的事务管理概念,研究进化仿真中的分区方法和事务管理机制;应分别在同步节点和异步节点的条件下,研究基于各种时间管理策略指导下的优化调度算法,及异常情况下的补偿措施,并探索动态的分区方案和负载均衡策略,乃至设计云模式下分布式仿真的部署方案;最后归纳出进化仿真研究的一般思路。
The main approach for artificial life research is by computer evolution simulation; its research results are applicative in areas such as multi-agent system, biochemical process simulation, military exercise, group behavior simulation, complex system, etc. The computational complexity of evolution simulation is usually very high. To summarize the characteristics, hot areas and aporias of this area, a comprehensive analysis, including classification, comparison and summarization, was made on related literatures using methods of bibliometrics, digesting, systematic review etc. The findings are: to address the issue of high computational complexity, considering the tighten space-time coupling of the simulation system, partition and transaction management must be utilized; both synchronous and asynchronous node structures should be considered when optimizing propelling strategies; dynamic partitioning and load balancing should be studied as well as the compensating strategy for exceptions; deployments for loose coupling environments or clouds can be one issue. A general research route is given.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第5期959-968,共10页
Journal of System Simulation
关键词
人工生命
进化仿真
多主体系统
分布式仿真
调度算法
artificial life
evolution simulation
multi-agent system
distributed simulation
scheduling algorithm