基于动态概率关系模型的规划识别研究及应用

The Research and Application of Plan Recognition Based on Dynamic Probabilistic Relational Models

根据Sanghai等人在概率关系模型的基础上引进的动态概率关系模型的概念,把该理论应用于规划识别的研究。动态概率关系模型是动态贝叶斯网络(DBNs)的一种扩展。通过引进动态概率关系(DPRM)讨论了动态不确定性问题。粒子滤波算法,作为在DBNs中的标准推理方法,在运用与DPRMs时有严重的局限性。利用Rao-Blackwellisation关系理论可以有效的改善,进一步再引入规划树概念,使DPRMs在机械装配领域得到了成功的应用。

Applying the theory of dynamic probabilistic relational model which introduced by Sanghai according to the basis of probabilistic models. PDRMs are an extension of Dynamic Bayesian networks（DBNs） ,The paper ad- dresses the problem of dynamic and uncertainty by introducing dynamic probabilistic relation model （DPRMs）. Particle filtering, the standard method for inference in DBNs, has severe limitations when applied to DPRMs, but a form of relational Rao-Blackwellisation could greatly improve its performance. Further reintroduces abstraction trees, making DPRMs successfully applying to the domain of assembly plan.