Generative AI For Automated Design of High-Intensity Interval Training Regimens: A Comparative Study Of Model-Based And Data-Driven Approaches

Abstract

High-Intensity Interval Training (HIIT) is widely recognized as an efficient exercise modality for improving cardiovascular fitness, metabolic health, and overall physical performance. However, the effectiveness and safety of HIIT depend heavily on individualized prescription of intensity, interval duration, recovery structure, and progression. Recent advances in Generative Artificial Intelligence (AI) offer new possibilities for automating and personalizing HIIT program design, yet systematic comparisons between different AI-driven approaches remain limited. This study presents a comparative framework examining model-based and data-driven generative AI approaches for automated HIIT regimen design. The model-based approach relies on physiological principles and constrained optimization to generate safe and interpretable training prescriptions, whereas the data-driven approach leverages wearable sensor data, machine learning, and reinforcement learning to adaptively personalize training sessions. A unified representation of HIIT regimens is proposed, integrating session-level structure and program-level progression while enforcing safety and physiological constraints. Comparative evaluation criteria include safety compliance, physiological training stimulus, personalization capability, and predicted adherence. The study highlights that while model-based approaches offer superior interpretability and safety assurance, data-driven generative models demonstrate greater adaptability and personalization potential. The findings suggest that hybrid systems combining generative AI with physiological constraint layers may represent the most promising direction for scalable, safe, and effective automated HIIT prescription. This research contributes to the growing field of AI-driven exercise prescription by offering a structured methodological framework suitable for digital health, sports science, and human-centered AI applications.