机器人室内语义建图中的场所感知方法综述

Place Perception for Robot Indoor Semantic Mapping:A Survey

场所感知问题是机器人语义地图研究的关键问题之一,本文对室内语义地图相关的场所感知方法进行全面综述.首先,根据近年的文献给出场所概念的描述性定义,对研究中涉及的相近术语和概念进行辨析,澄清研究对象和研究主题.然后,根据实现场所感知目标所采用的线索对已有方法进行分类介绍.主要分成3个大类:基于环境布局几何信息的方法、基于环境布局视觉信息的方法、基于用户指导信息的方法,其中各类又根据所用信息特点细分为若干子类.除此之外,将一些特殊研究方法单独归类进行补充说明.阐述各类别方法对场所感知问题的解决思路和工作原理,并指出各种方法特点和局限性.最后,分析了该领域存在的主要问题,并对未来研究方向进行了讨论和展望.

Place perception is a key problem in the robot semantic map research area, and a comprehensive survey of place perception related to indoor semantic mapping is presented in this paper. Firstly, a descriptive definition of place concept is given according to the latest research literature. On the other hand, some similar terms and concepts involved in this research are discriminated in order to clarify the research object and theme. Then, the existing methods are classified in terms of the clues which are used to implement the place perception targets. There are mainly three categories： the methods based on geometry information of environmental layout, the methods based on visual information of environmental layout, and the methods based on user guiding information. Each category can be further divided into some sub-categories according to the characteristics of used information. Besides, some particular methods are also described as an additional category. The ideas and principles used in all methods for solving place perception problem are comprehensively introduced, and the characteristics and limitations of these methods are also pointed out. Finally, some potential issues are analyzed, and probable future research directions are discussed.