Creating Lifelike Artificial Social Agents: The Role of Movement Strategies in Virtual Realism

Abstract

This thesis investigates the realism of human movement in Artificial Social Agents (ASAs) within virtual environments, aiming to enhance the authenticity and engagement of human-ASA interactions. Addressing a critical gap in existing literature, this research focuses on the nuanced micro-movement strategies and natural navigational behaviours essential for creating lifelike virtual agents. By integrating psychological theories, advanced technological developments, and various distance and orientation classifications, the study presents a comprehensive framework for improving ASA realism. The methodologies employed include extensive recording and analysis of human locomotion data, which serve as the basis for developing sophisticated movement strategies for ASAs. This data-driven approach ensures that the proposed strategies closely mimic the subtle, unconscious micro-movements that characterize natural human behaviour. Additionally, the perceived realism of these movements is evaluated through rigorous user studies, providing empirical evidence on the effectiveness of the proposed movement strategies. The findings underscore the significance of adaptive, context-aware behaviours in ASAs, demonstrating that the incorporation of sophisticated algorithms and machine learning techniques can substantially enhance the realism and engagement of virtual interactions. This research has broad implications across multiple domains, including education, healthcare, and social robotics, where the authenticity of virtual agents is pivotal in enhancing user experience and efficacy. The insights gained from this study pave the way for future advancements in the design and implementation of more lifelike and responsive ASAs, ultimately contributing to more realistic and effective virtual environments.