Open-source AI base for humanoid with kinect sensor in education context

Abstract

This diploma thesis explores the communication gaps in Social Assistive Robots (SARs), with a particular focus on their application in educational settings. The project aims to address these significant gaps by incorporating the ability to use human body language in a humanoid robot, focusing on the difficulty robots encounter in replicating subtle human non-verbal cues essential for effective communication. During the building of a historical review, the study underscores how technological advancements, especially the entry of artificial intelligence (AI), have evolved through the Industrial Revolutions, shaping modern robotics and the process of humanoids. The relationship between these historical milestones and the ongoing progress in AI-driven robotics is examined, emphasizing their impact on improving efficiency. Furthermore, this research introduces precision into various domains, particularly in the context of supporting children with Autism Spectrum Disorder (ASD). This free and open-source integration endeavors to establish an accessible and effective solution. This implementation specifically addresses the limitations of closed and proprietary software, often exclusively supported by their respective manufacturers. Consequently, this research highlights the potential for older products to find renewed utility within the evolving technological landscape. The thesis delineates the design and development of a prototype system aimed at enhancing human-robot interaction. In addition, the design and implementation of an AI base tailored for educational purposes is presented, with a system that presents a base with combination of the Microsoft Kinect, Raspberry Pi 4, a Laptop computer and the virtual environment of NAO. While the testing and evaluation highlights the successful implementation of various system components, addressing challenges such as the limitations imposed by software and hardware levels. Consequently, the identified limitations constrain the scope of the application. As such, the troubleshooting process comprehensively analyzes all potential avenues towards a viable solution. Moreover, the progress of the testing and evaluation provides modifications to complete the application on the function layer. Also the key achievements such as the successful connection of MariaDB with MQTT, wireless communication between MQTT and the NAO robot, and the management of Kinect captures are noted. Despite all opposition, progression is also used to test system performance and simulate different scenarios in a virtual NAO environment. Moreover, it proposes potential future extensions of the system on overcoming existing limitations and further strengthening the system's applicability in educational and therapeutic settings in real-time.