Study on immersive simulation and effect evaluationof sports training combined with virtual reality technology

Traditional approaches to sports training often rely on subjective coaching feedback, limited biomechanical measurements,and coarse performance evaluations that fail to capture the intricate multi-scale temporal and spatial dynamics of athleticmovements. These conventional methods struggle with the high variability in individual athlete biomechanics, complexinter-joint dependencies, and cross-context adaptation challenges, limiting their effectiveness in providing precise, adaptive,and scalable training assessments. To address these limitations, we introduce a novel computational framework combining theHierarchically Coupled Spatio-Temporal Dynamics Network (HC-STDN) with the Adaptive Domain-Regularized OptimizationFramework (ADROF). The HC-STDN leverages hierarchical decomposition of skeletal motion, multi-scale temporal dynamicsextraction, spatio-structural relational graph processing, and attention-based temporal aggregation to robustly modelthe intricate motion patterns inherent in sports activities. Complementarily, ADROF integrates supervised learning withdomain-adaptiveization, biomechanical consistency constraints, multi-scale temporal smoothness, and Wasserstein-baseddistribution alignment, further enhanced by instance-level adaptive weighting and auxiliary sub-task supervision. Our unifiedmethodology demonstrates clear superiority over current techniques in a wide range of sports contexts, delivering enhancedprecision, resilience, and adaptability in assessing athlete performance across multiple domains, environmental conditions,and sensing technologies. This comprehensive solution exemplifies the convergence of immersive simulation, interactivemachine intelligence, and data-driven biomechanical modeling, contributing directly to the interdisciplinary themes prioritizedby Frontiers in Computer Science.