Online Engagement Detection and Task Adaptation in a Virtual Reality Based Driving Simulator for Autism Intervention

摘要

Individuals with Autism spectrum disorder (ASD) have difficulty functioning independently on essential tasks that require adaptive skills such as driving. Recently, computer-aided technology, such as Virtual Reality (VR), is being widely used in ASD intervention to teach basic skills to children with autism. However, most of these works either do not use feedback or solely use performance feedback from the participant for system adaptation. This paper introduces a physiology-based task adaptation mechanism in a virtual environment for driving skill training. The difficulty of the driving task was autonomously adjusted based on the participant's performance and engagement level to provide the participant with an optimal level of challenge. The engagement level was detected using an affective model which was developed based on our previous experimental data and a therapist's ratings. We believe that this physiology-based adaptive mechanism can be useful in teaching driving skills to adolescents with ASD.