Quality of experience and depth perception in spatial augmented reality games
Milad Ghanbari · University of Klagenfurt
Accurate depth perception is an important factor for realistic interaction in Augmented Reality (AR), especially for physical tasks that require precision, such as throwing. This thesis measures how judging virtual depth affects user performance by using SDART (Spatial Dart AR Simulation with Hand-Tracked Input), a system developed for the Apple Vision Pro (AVP). The simulation was created with the Unity engine, using the PolySpatial SDK for visionOS integration and ARKit to process the AVP’s Six-Degrees-of-Freedom (6DoF) controller-free hand-tracking data. The main research gap addressed is the limited empirical evidence on how users perceive and physically adjust to different virtual distances in a high-quality AR environment without traditional controllers. A mixed-methods user study with 22 participants was conducted. The experiment used a counterbalanced, repeated-measures design to reduce learning effects. The main independent variable was throwing distance, with participants performing tasks at three different target depths: Near (1.87m), Standard (2.37m), and Far (2.87m). In addition, a time-pressure condition was included, consisting of a 60-second time trial at the standard distance, to examine performance under increased cognitive demand. Objective performance data, including score, hit positions, and timing for each throw, were automatically recorded in CSV files. After completing the tasks, participants provided subjective feedback through a detailed Likert-scale questionnaire that focused on perceived depth, control, realism, and comfort.The analysis of the objective data showed a statistically significant decrease in performance as target distance increased, clearly showing the effect of depth perception on user accuracy. During the time-trial condition, participants threw much faster, with an average reduction of about 25% in time per throw. However, unlike a typical speed-accuracy trade-off, scoring accuracy did not decrease, which suggests that users were able to adjust their throwing speed without losing precision. In addition, subjective ratings for “It was easy to judge the distance” and “I felt in control” stayed high across participants, even for those with lower objective performance. This result suggests that the interaction quality of the system supports a strong sense of control and immersion, regardless of task difficulty or individual skill. Overall, this research offers a clear quantitative approach to studying how spatial and cognitive limits affect interaction in AR and confirms the AVP’s ability to provide effective and intuitive spatial cues for complex, physics-based tasks.