Creating conceptual maps by natural language descriptions for human - friendly robot navigation

Abstract

The living standards of elderly people are uplifted through developing assistive robots who are capable of supporting their daily activities by possessing the competency to provide companionship to human beings. Human activities are frequently related to navigational tasks and human tend to use descriptions which include natural language phrases and the terms which describe the distance related uncertainties such as little, large, near, far, small, close to describe about spatial information. Therefore assistive robots should be capable of analyzing and understanding descriptions which include natural language phrases. The best and foremost navigation skills of a robot are repre- sented with the competency of virtual imagination related to an unknown environment. Subsequently the quality of virtual imagination of the robot should be improved with the experiences as human beings. The quality of understanding the user appropriately, e ciently and e ectively plays a vital role in order to expand their knowledge as well as the experiences collected from day to day performances and conversations. Accord- ingly, the robot should contain a memory and an advance knowledge base including the information regarding objects that it experiences in day to day activities. The require- ment of a user should be clearly con gured with the capability of processing data in order to grasp the relationships among attributes of objects. Obviously, a robot should possess the competency to interpret spatial information in the mode of uncertain terms. Signi cantly the robot should be capable enough to enhance the knowledge through e ective communication with the user. This research proposes a procedure to understand spatial information in a description with the uncertain terms and creates a conceptual map in a robot memory which can be linked with spatial cognitive map for purposeful, e ective and human friendly naviga- tion task. The proposed method is consisted with creation of cognitive object maps and cognitive spatial maps. Further, both maps are created based on information conveyed through interactive conversations between the user, robot and the vocal descriptions. The conceptual maps are created by amalgamating the spatial and object cognitive maps. Moreover, the conceptual map also can be updated using the conversations occurred in the interactions among robot and the user. Furthermore, this research pro- poses a procedure to enhance the capability of virtual imagination of a service robot while understanding the information regarding the uncertainties of an object size with the aid of arti cial neural network. In addition to that gesture command identi cation method and vocal navigation command identi cation method are also implemented. Moreover, navigation commands are categorized and studied for the unique attributes such as uncertainty, incompleteness, inconsistency and unpredictability to improve hu- man robot interaction. The proposed method is validated by the experiments using MI Rob platform (Version 2.0). Further, a software platform was introduced by integrating the implementation of proposed methods. Software platform is consisted of processing spatial maps for room boundary establishment to object identi cation considering spatial information acquired through descriptions and conversations. System generated data was compared with data derived from human studies and analyzed using statistical methods in order to validate system accuracy. The proposed system can be used to enhance the human- robot interactions and to perform navigational tasks more e ectively and purposefully in a previously unknown environment.