基于承担特质模型的机器人动作序列生成方法  

An Action Sequence Generation Method for Nuclear Power Plant Robot Based on Affordance Model

作　　者：刘春阳[1] 张敬伟[2] 郑雪峰[1] 涂序彦[1] 

机构地区：[1]北京科技大学计算机与通信工程学院,北京100083 [2]华北电力大学控制与计算机工程学院,北京102206

出　　处：《计算机仿真》2014年第7期334-337,411,共5页Computer Simulation

基　　金：新世纪优秀人才支持计划(NCET-11-0634);自然科学(青年)基金(61105083);中央高校基本科研业务费专项资金资助(12ZX16;GrantNo.11QG12)

摘　　要：针对核电站检修机器人所搭载的机械手臂在执行作业任务过程中,如何得到一个操作任务的基元动作序列这一关键问题,提出了一种承担特质模型与贝叶斯网络相结合的推理方法。首先,根据承担特质模型分析构建机器人操作的贝叶斯网络所需的节点,由经验数据搭建初始的贝叶斯网络;然后,机器人在此基础上通过与周围环境的交互实践进行学习,提供完善初始贝叶斯网络所需的学习数据,以弥补构建初始贝叶斯网络时人为给定经验数据的不足;最后,通过检验动作集合的边缘分布值,来验证提出的方法在动作序列生成问题中的可行性和有效性。The key issue of the process for nuclear power plant maintenance robot equipped with a mechanical arm to perform job tasks, is how to get operational task primitive action sequences. To solve this problem, this paper proposed an inference method combining affordance model with Bayesian network. Firstly, the nodes needed by the Bayesian network were analyzed based on the affordance model, with the Bayesian network therein built by empirical data. And then, the required learning data for the improvement of initial Bayesian network were thus obtained to ful- fill the compensation for shortage of the empirical data, which was utilized to build initial Bayesian network. At last, this method was proved to be effective and feasible when undergoing a series of experiments.

关 键 词：机器人 承担特质模型 贝叶斯网络 动作序列 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]