Skip to main content

Robotics and Autonomous Systems Lab
at Vanderbilt

Physiology-based Dynamic Difficulty Adjustment Driving Task

Personnel: Dayi Bian, Joshua Wade, Zachary Warren, & Nilanjan Sarkar.

Goals/Objectives:

  1. Design a physiology-based dynamic difficulty adjustment driving task to keep the user optimally challenged

Outline:

VR-based skill training systems have demonstrated usefulness for training specific skills in individuals with ASD. However, a simple performance-based system without appropriate feedback to the user may not be optimal for enhancing and optimizing learning. Managing task difficulty to keep the user optimally challenged is crucial for designing an effective skill training system. In this project, random forest algorithm was used to build an engagement prediction model based on the physiological responses (blood volume pulse, skin conductance and respiration) from the users. The model was used to predict the user’s engagement model during the driving task. Then the driving task adjusts its difficulty level based on the user’s performance metrics and engagement level in real time.

Publications:

  1. D. Bian, J. Wade, A. Swanson, A. Weitlauf, Z. Warren, and N. Sarkar, “Online Engagement Detection and Task Adaptation in a Virtual Reality Based Driving Simulator for Autism Intervention,” International Conference on Universal Access in Human-Computer Interaction, 2016.
  2. D. Bian, J. Wade, A. Swanson, A. Weitlauf, Z. Warren, and N. Sarkar, “Design of a Physiology-based Dynamic Difficulty Adjustment Virtual Reality Driving Platform and its Application in ASD Intervention,” ACM Transactions on Interactive Intelligent Systems (TiiS), 2017, Under review.