预测状态表示综述

Survey of predictive state representations

预测状态表示是描述离散时间有限状态的动态系统的新方法。使用动作—观测值序列的预测向量表示系统状态在将来时刻发生的概率,能解决现有动态系统决策过程中计算复杂的问题。综述了预测状态表示的基本原理,介绍了预测状态表示的建模过程和规划算法,对已有的建模方法和规划方法进行总结分析和比较,指出了该研究领域的发展方向,最后提出了研究面临的挑战。

Predictive state representations （ PSRs ） are new models for discrete-time finite action and observation stochastic systems. Because a PSR represents the system＇ s state as a set of predictions of the observable outcomes of tests performed in the system, it can solve the computing problems in exist stochastic decision systems. This paper introduced the principles of PSR models, surveyed the PSR model and planning techniques, analyzed and compared the fundamental principles behind the modeling and planning algorithms of PSR, pointed out the development trend, and gave the challenges that the research of PSR was facing.