一种面向室内智能机器人导航的路径自然语言处理方法  

A Route Instruction Method Using Natural Language Processing for Indoor Intelligent Robot Navigation

作　　者：李新德[1] 张秀龙[1] 

机构地区：[1]东南大学自动化学院复杂工程测量与控制教育部重点实验室,南京210096

出　　处：《自动化学报》2014年第2期289-305,共17页Acta Automatica Sinica

基　　金：国家自然科学基金(61175091);江苏省自然科学基金(BK2010403);东南大学优秀青年教师教学;科研资助计划(3208001203);江苏省"青蓝工程"资助计划资助~~

摘　　要：为了实现通过自然语言指导机器人导航的目的,提出了一种利用描述路径的自然语言直接绘出机器人运行路径的方法.为了提高机器人对路径自然语言的理解,采用了基于组块的分析方法.首先深入地研究了收集到的路径描述的自然语言语料库,总结出了语义和句法的对应关系.在此基础上,构造了层叠条件随机场,通过采用组块分析方法,先进行名词短语组块的提取,然后提取语义组块.为了解决名词短语的解析问题,引入了名词实体关系推理方法.并且在提取出的语义组块的基础上逐个建立路径单元,组成完整的路径信息.最后,给出了自然语言处理各个环节的实验结果,并结合具体语句展示了从文字到机器人导航地图的映射过程,对开展人机交互式和谐导航具有重要意义.In this paper, a new route instruction method using natural language processing was investigated to guide mobile robot navigation. To gain a clear understanding of the route expressed by natural language, a chunk parsing method was adopted. Firstly, the collected corpus of the route expressed by natural language was deeply studied and the corresponding relationship between the sentence structure and semantics was presented. Based on this, cascaded conditional random fields were constructed and the noun phrase chunk is picked up and, its semantics chunk was also picked up by applying the chunk parsing method. To resolve the meaning of the noun phrases, a noun entities relationship reasoning method was presented. The units of the route was built with the picked-up semantic chunks and an integral route was finally completed. At last, the experimental results of every loop were presented. Some example sentences of the route expressed by natural language were given and processed to show the procedure from the sentences to map used in the robot navigation. It is very significant for robot navigation with harmony based on human-robot interaction to carry out this research.

关 键 词：自然语言处理 机器人导航 层叠条件随机场 贝叶斯推理 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]