一种优化的一致性规划状态变量选择算法

研究了一致性规划任务信念状态空间的表示方法。针对一致性有限域表示(CPT-FDR)算法在任务生成阶段选择状态变量的不足，提出了一种基于初始状态中文字相容互斥的状态变量选择算法——MECV算法。CPT-FDR未考虑初始信念状态中文字的互斥性，产生冗余的编码信息，降低了编码的效率。MECV算法利用有用正负文字构造新的未覆盖事实集，提取初始信念状态中处于不同世界状态的文字组成互斥组，再编码状态变量。实验结果表明该算法能有效地压缩信念状态空间。

An Optimized State Variables Selection Algorithm in conformant planning Task

This study focused on the representation of the belief states in conformant planning tasks. The state variables selection algorithm in the process of CPT-FDR generation is not enough to deal with the mutex conditions. A new method named MECV (mutually-exclusive-choose-variables) was put forward. CPT-FDR does not think over the mutually-exclusive between initial belief states. It generates redundant information which reduced the efficiency of encoding. First, MECV used the useful positive and negative literal to make a new uncovered fact set. And then it took out the mutually exclusive literal from different real state which in the initial belief states to make a mutually-exclusive-group. Finally, it used the new uncovered fact set and mutually-exclusive-group encoding state variables. The outcomes of comparative experiments validate the effectiveness of the new algorithm.