Wearable Biosensing and Machine Learning for Data-Driven Training and Coaching Support.

Area of Science:

• Sports Science and Technology
• Biomedical Engineering
• Data Science in Athletics

Background:

• Wearable biosensing systems are increasingly used in sports for continuous monitoring and data-driven training.
• The practical utility of these systems for coaching hinges on data validity, model robustness, and empirical evaluation conditions.

Purpose of the Study:

• To review empirical evidence on wearable biosensing, signal processing, and AI/ML-based adaptive training systems.
• To construct an evidence map for transparency, detailing study parameters and outcomes.

Main Methods:

• Structured narrative review of Scopus, PubMed, Web of Science, and Google Scholar (2010-2026).
• Synthesis of empirical and applied evidence on biosensing and ML adaptive training.
• Creation of an evidence map summarizing sensing modalities, cohorts, settings, models, and evaluations.

Main Results:

• Wearable biosensors reliably capture physiological, biomechanical, and biochemical markers with proper quality control.
• ML models improve training adaptation and recovery estimation over traditional metrics in various sports.
• Limitations include motion artifacts, variability, lab-field shifts, and missing contextual data.

Conclusions:

• AI-driven biosensor systems are valuable decision-support tools for monitoring and adaptive training.
• Effectiveness depends on sensor reliability, validation, and human oversight; they are not autonomous coaching agents.
• Human-AI collaboration, where AI aids interpretation and coaches provide context, is the most effective model.