Avatar Control using Eye-Tracking in Virtual Reality : A user study on the sense of embodiment, user experience and workload using a Unity3D Prototype

Abstract: This master’s thesis investigates methods for improving accessibility in virtual reality (VR), focusing on avatar control for individuals with limited mobility. VR offers an opportunity to help them with specific rehabilitation. Avatars, representing the user, serve as a bridge between the virtual and physical world, helping with the sense of embodiment and the more control, the better is the rehabilitation, for example. The objective is to identify effective avatar control approaches in VR, excluding traditional body movements, and assess their impact based on four criteria: sense of embodiment, user experience, task workload, and performance. To address this, we designed a VR experiment in Unity3D with two tasks: one for interaction (selection and manipulation) and one for navigation. Each task had two independent variables. One common variable was the input modality (head-tracking or eye-tracking). For interaction, the second variable determined whether participants had to grasp an object to trigger the avatar’s reaction or could directly control the avatar. In navigation, the second variable decided whether participants used steering or triggering mechanisms for navigating towards their destination. A user study using an HTC Vive Pro eye including 24 healthy participants, testing all four conditions for both tasks was conducted. Participants received tutorials before each condition. We recorded run times during each condition, with five runs for interaction and four for navigation. After each condition, participants provided feedback: the Virtual Embodiment Questionnaire assessed sense of embodiment, the NASA Task-Load Index evaluated task workload, and the System Usability Scale gauged user experience. A repeated-measure two-way Anova showed that for interaction, eye-tracking improved speed performance but did not significantly affect the other metrics compared to head-tracking. In navigation, head-tracking consistently outperformed eye-tracking in all aspects. No other significant results were found. In conclusion, this thesis lays a foundation for enhancing VR accessibility by providing insights into input modalities and avatar control methods.