Personalized and Adaptive Virtual Reality Training for Physically Coupled Robots in Construction

Abstract

Work-related musculoskeletal disorders (WMSDs) significantly impact worker health and productivity in construction.Physically coupled robotics, such as exoskeletons, offer a promising solution by augmenting human capabilities and reducing strain.However, their effective use requires precise humanrobot synchronization, which can be challenging in dynamic construction environments.Traditional training methods often fail to address these complexities, leading to suboptimal performance, misuse, or increased injury risk.To overcome these challenges, this study develops a personalized Virtual Reality (VR) training platform incorporating an Individualized Bayesian Knowledge Tracing (iBKT) framework.The system continuously evaluates user performance in key body regions, waist, knee, and shoulder, using real-time metrics to dynamically adjust task difficulty, training frequency, and instructional feedback.This adaptive approach tailors training content to enhance motor coordination and postural control in exoskeletonassisted tasks.Results indicate that ATG achieved greater reductions in postural errors across tasks, with total ER decreasing by 6, 3, and 5 points for walking, bending, and squatting, respectively, compared to 4, 2, and 5 points in TTG.These findings underscore the importance of adaptive, personalized training in enhancing human-robot interaction, safety, and productivity.By equipping workers with tailored skills, this study advances workforce-robot collaboration, fostering more resilient and efficient construction practices.