Artificial self-awareness for robots

AlQallaf, Ali H.H.A.H. (2023) Artificial self-awareness for robots. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 2022AlQallafPhD.pdf] PDF
Download (82MB)


Robots are evolving and entering into various sectors and aspects of life. While humans are aware of their bodies and capabilities, which help them work on a task in different environments, robots are not. This thesis is about defining and developing a robotic artificial self-awareness framework. The aim is to allow robots to adapt to their environment and better manage their task. The robot’s artificial self-aware knowledge is captured based on levels where each level helps a robot acquire higher self-awareness competence. These levels are inspired by Rochat [1] self-awareness development levels in humans, where each level is associated with a complexity of self-knowledge. Self-awareness in humans leads to distinguishing themselves from the environment, allowing humans to understand themselves and control their capabilities. This work focuses on the first and second levels of self awareness through differentiation and situation (minimal self).

The artificial self-awareness level-1 proposes the first step towards a basic, minimal self-awareness in a robot. The artificial self-awareness level-2 proposes an increasing capacity of self-awareness knowledge in the robot. That is, this thesis posits an experimental methodology to evaluate whether the robot can differentiate and situate itself from the environment and to test whether artificial self-awareness level-1 and level-2 increase a robot’s self-certainty in an unseen environment.

The research utilises deep neural network techniques to allow a dual-arm robot to identify itself within different environments. The robot vision and proprioception are captured using a camera and robot sensors to build a model that allows a robot to differentiate itself from the environment. The level-1 results indicate that a robot can distinguish itself with an accuracy of 80.3% on average in different environmental settings and under confounding input signals. Also, the level-2 results show that a robot can situate itself in different environments with an accuracy of 86.01% yielding a higher artificial self-certainty of 5.71%. This thesis work helps a robot be aware of itself in different environments.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Colleges/Schools: College of Science and Engineering > School of Computing Science
Supervisor's Name: Aragon Camarasa, Dr. Gerardo
Date of Award: 2023
Depositing User: Theses Team
Unique ID: glathesis:2023-83422
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 10 Feb 2023 11:48
Last Modified: 10 Feb 2023 11:52
Thesis DOI: 10.5525/gla.thesis.83422
Related URLs:

Actions (login required)

View Item View Item


Downloads per month over past year