论文部分内容阅读
交互式动态影响图(I-DIDs)是基于概率图形理论的多智能体动态交互决策的图模型.为缓解该模型状态空间随时间片增加呈指数级增长的趋势,文中基于行为等价的基本思想压缩状态空间,提出构建Epsilon行为等价类的方法:利用有向无环图表示其它Agent可能的信度和行为,把信度在空间上接近的模型聚为一类,实现自顶向下合并行为等价模型.该过程避免求解状态空间中的所有候选模型,节省了存储空间和计算时间.模型实例上的仿真结果显示了该算法的有效性.
Interactive dynamic impact graph (I-DIDs) is a graph model of multi-agent dynamic interactive decision making based on probability graph theory.In order to alleviate the trend of exponential growth of state space of this model, Thought compression state space, a method to construct Epsilon’s behavior equivalence class is put forward: using the directed acyclic graph to represent the possible reliability and behavior of other Agents, the model of reliability close to each other in space is grouped into a category, This method avoids solving all the candidate models in the state space and saves storage space and computing time.The simulation results on the model examples show the effectiveness of the algorithm.