论文部分内容阅读
针对部分可观察马尔可夫决策过程(POMDP)的信念状态空间规模“维数灾”问题,根据信念状态变量存在可分解和独立关系的特性,提出一种基于动态贝叶斯网络(DBN)的可分解信念状态空间压缩算法(factoredbelief states space compression,FBSSC).该算法通过构建变量间依赖关系图,根据独立关系检验去除多余边,将转移函数联合概率分解成若干个条件概率的乘积,实现信念状态空间的无损压缩.对比实验和RoboCupRescue仿真结果表明,本文算法具有较低误差率、较高收敛性和普遍适用性等特性.
According to the existence of the decomposable and independent relationship between the state variables of beliefs and the believable state space scale “dimensionality disaster ” of partially observable Markov decision process (POMDP), a dynamic Bayesian network (DBN (FBSSC) .This algorithm constructs the dependency graph between variables, removes redundant edges according to the independent relationship test, decomposes the joint probability of transfer function into the product of several conditional probabilities, The lossless compression of the belief state space is achieved.Comparison experiments and RoboCupRescue simulation results show that the proposed algorithm has the characteristics of lower error rate, higher convergence and universal applicability.