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.
Citation:
Bandara, H.M.R.T. (2019). Creating conceptual maps by natural language descriptions for human - friendly robot navigation [Master's theses, University of Moratuwa]. Institutional Repository University of Moratuwa. http://dl.lib.uom.lk/handle/123/18632