This study addresses persistent challenges in youth soccer training, including limited personalization, delayed scientific feedback, and the lack of injury risk early warnings. To fill gaps in applicability, functional completeness, and technological integration, an innovative training optimization framework based on an Artificial Intelligence (AI) digital twins network was developed. The framework integrates publicly available large-scale match and training datasets, Wyscout and StatsBomb, to create virtual twins. These twins map athletes’ multidimensional states—including position, actions, and physiological signals—in real time. Deep learning was applied for high-precision feature extraction and movement trajectory tracking. Reinforcement learning mechanisms enabled the system to adaptively optimize training strategies and load based on real-time feedback. Experimental results demonstrate that the model significantly outperforms traditional methods. Real-time player position tracking reached 95% accuracy. Motion pattern analysis and classification efficiency improved by 20%. The system also identifies potential high-risk movement patterns, providing early warnings to prevent injuries. In summary, the framework achieves a closed loop from data perception to intelligent decision-making and provides coaches with an actionable, personalized, and evidence-based tool for scientific training. This approach has important theoretical and practical implications for improving the quality, safety, and long-term development of youth soccer training.
Han et al. (Wed,) studied this question.