Joint Function and Movement Variability During Daily Living Activities Performed Throughout the Home Setting: A Digital Twin Modeling Study

Human mobility is commonly assessed in the laboratory environment, but accurate and robust joint motion measurement and task classification in the home setting are rarely undertaken. This study aimed to develop a digital twin model of a home to measure, visualize, and classify joint motion during activities of daily living. A fully furnished single-bedroom apartment was digitally reconstructed using 3D photogrammetry. Ten healthy adults performed 19 activities of daily living over a 2 h period throughout the apartment. Each participant's upper and lower limb joint motion was measured using inertial measurement units, and body spatial location was measured using an ultra-wide band sensor, registered to the digital home model. Supervised machine learning classified tasks with a mean 82.3% accuracy. Hair combing involved the highest range of shoulder elevation (124.2 ± 21.2°), while sit-to-stand exhibited both the largest hip flexion (75.7 ± 10.3°) and knee flexion (91.8 ± 8.6°). Joint motion varied from room to room, even for a given task. For example, subjects walked fastest in the living room (1.0 ± 0.2 m/s) and slowest in the bathroom (0.78 ± 0.10 m/s), while the mean maximum ankle dorsiflexion in the living room was significantly higher than that in the bathroom (mean difference: 4.9°, <i>p</i> = 0.002, Cohen's d = 1.25). This study highlights the dependency of both upper and lower limb joint motion during activities of daily living on the internal home environment. The digital twin modeling framework reported may be useful in planning home-based rehabilitation, remote monitoring, and for interior design and ergonomics.