Attention-Safe Augmented Reality System with Edge Computing
Zhehan Qu · 2024
While created to be helpful and effective, augmented reality (AR) applications can be task-detrimental. In particular, head-mounted AR can have a negative impact on user attention that is difficult to notice or avoid due to its dominance over the entirity of the user’s field of view. This position paper proposes to investigate how we can detect and mitigate suboptimal attentional states and how those contexts can help develop a system that ensure attention-safe AR experience. To address these questions, we utilize the eye-tracking capability of existing commercial AR headsets and propose to develop a real-time edge computing system that runs machine learning (ML) algorithms to detect and mitigate task-detrimental user attention in AR. Based on seconds of eye-tracking data collected from users participating in preliminary user studies using our customized AR app, ML models were able to predict distractions happening in AR, users’ attention control ability, and task performance. We are currently working on enhancing the generalizability of our models on different AR applications and user groups, and developing an efficient edge computing system that can provide real-time feedback to AR applications. We envision that combining our proposed system with existing AR frameworks can help create safe and effective AR experiences that can adapt to individual user attention patterns.