摘要
针对多Agent系统(MAS)内外环境变化所产生的不确定性和任务分配序列决策的要求,利用马尔科夫决策过程(MDP)模型对MAS中的动态任务分配问题进行了分析和建模.其中,状态空间由各Agent的当前负载和待分配的任务组成,每一状态下有多种任务分配方案,利用迭代方法可以获得最佳的任务分配方案以实现系统长期收益最大化的目标.仿真实验表明,MDP模型可以合理地模拟MAS中任务分配的运作过程,并在小规模环境下方便地获取最优任务分配策略.
According to the environmental uncertainty in multi-agent system (MAS) and the requirement of sequential decision-making, task allocation is analyzed and modeled with Markov decision process(MDP). The state space consists of agents' current loads and the allocating tasks. In each state, there are many strategies to allocate tasks to agents. Iterative method is used to get the best allocation to maximize the system benefits in long time, It is proved that MDP model can simulate the process of task allocation to get the optimal task allocation strategy.
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2007年第1期54-57,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(60274065
70572034).