Locability: An Ability-Based Ranking Model for Virtual Reality Locomotion Techniques

2025-10-07

AI Summary

Over one hundred locomotion techniques in VR exhibit heterogeneous compatibility with users’ upper-limb motor abilities due to variations in control modalities (e.g., gestures, button presses, body motions) and the heterogeneity of upper-limb impairments. Method: We propose the first capability-oriented locomotion technique ranking model, integrating user capability assessments (via a standardized questionnaire) and fine-grained interaction behavioral data, and employing machine learning to predict individual performance across six mainstream locomotion techniques and rank them accordingly. Contribution/Results: Evaluated with empirical data from 20 participants—both with and without upper-limb impairments—the model achieves 92% accuracy in identifying the optimal technique per user and predicts navigation time with only 12% mean absolute error. For global ranking across all six techniques, it attains 61% top-1 accuracy and ≤8% time prediction error. This work establishes the first capability-driven, personalized locomotion adaptation framework, significantly advancing VR accessibility and inclusive design.

Technology Category

• Humans and AI: Interaction Techniques and Devices
• Machine Learning: Learning Preferences or Rankings
• Computer Vision: Motion & Tracking

Application Category

• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

There are over a hundred virtual reality (VR) locomotion techniques that exist today, with new ones being designed as VR technology evolves. The different ways of controlling locomotion techniques (e.g., gestures, button inputs, body movements), along with the diversity of upper-body motor impairments, can make it difficult for a user to know which locomotion technique is best suited to their particular abilities. Moreover, trial-and-error can be difficult, time-consuming, and costly. Using machine learning techniques and data from 20 people with and without upper-body motor impairments, we developed a modeling approach to predict a ranked list of a user's fastest techniques based on questionnaire and interaction data. We found that a user's fastest technique could be predicted based on interaction data with 92% accuracy and that predicted locomotion times were within 12% of observed times. The model we trained could also rank six locomotion techniques based on speed with 61% accuracy and that predictions were within 8% of observed times. Our findings contribute to growing research in VR accessibility by taking an ability-based design approach to adapt systems to users' abilities.

Problem

• Predicting optimal VR locomotion for motor impairments
• Ranking techniques using ability-based machine learning
• Reducing trial-and-error in VR accessibility adaptation

Innovation

• Machine learning predicts fastest VR locomotion techniques
• Model ranks techniques based on user ability data
• Ability-based design adapts systems to individual capabilities