Enhancing XR Accessibility through Anthropometrically Diverse Biomechanical Simulation

Biomechanical user simulations can complement efforts to enhance XR accessibility by enabling rapid evaluation of alternative design choices across a wider range of user body types and capabilities. However, it is unclear to what extent current biomechanical simulation policies generalize to different body shapes. To address this gap, we investigate how biomechanical models for mid-air selection perform under anticipated diversity in user anthropometry. Using a musculoskeletal simulator with reinforcement learning (RL), we analyzed how variations in arm length influence model performance. The results reveal that existing RL policies are applicable only to the narrow range of anthropometric conditions for which they were trained, with significant performance deterioration outside this range. Our study highlights the importance of developing flexible and general biomechanical models capable of representing anthropometrically diverse users.