摘要
为了构造和部署大规模的多agent系统,人们必须找到并解决其基本问题,其中之一就是可能存在的局部性系统故障。这也就意味着,容错对于大规模多agent系统来说,是一个无法回避的主题。文中讨论了这类问题并且提出了一种多agent系统的容错方法。最先的想法是将复制策略运用到agent中,对处于危急状态的agent进行复制从而避免系统故障,但是由于agent的危急性会在执行过程中演变,并且agent的可用资源是绑定的,所以需要动态以及自动地调整agent的复制体个数,从而最大化它们的作用和可靠性。文中将描述评估某个agent危险性的方法以及相关机制,并且决定使用何种策略(如:主动复制,被动复制)以及如何将其参数化(如:复制的个数)。
To construct and deploy large - scale multi - agent system, must address and solve some fundarnental issues, one of which is the possibility of partial failures. This means that fault tolerance is an inevitable topic for the large - scale multi - agent system. So, in this paper,a new approach of fault tolerance in multi - agent systems is discussed. First introduce the notion of replication to the agent system and replicate the critical agents to protect the system from the failure. However, as the agent criticality will evolve in the course of executing and the available resources are bounded to the certain agents,need to dynamically adapt the number of replica agent in order to maximize their reliability and availability. Meanwhile, this paper will include the approach and mechanism to evaluate the agent criticality, how to decide the replica strategy (active or passive) and how to parameter them (numbers of replicas).
出处
《计算机技术与发展》
2007年第8期140-143,147,共5页
Computer Technology and Development
关键词
容错
多AGENT系统
复制
自适应
相互依赖
危险性
fault - tolerance
multi - agent system
replica
adaptive
interdependence
criticality