Model Predictive Control of Physical Activity for Long-Term Diabetes Risk Reduction

Abstract

This paper presents an output-feedback Model Predictive Control (MPC) framework designed to support long-term diabetes prevention through personalized physical activity planning. A physiological glucose–insulin dynamic model is used to predict how the human body responds to variations in physical activity. In this approach, physical activity is treated as a controllable input that can influence glucose regulation. Since continuous measurement of all physiological states is not feasible in practical healthcare settings, an output-feedback strategy combined with a state estimation technique is implemented to estimate unmeasured variables. The proposed controller dynamically adjusts the intensity, duration, and timing of physical activity to keep glucose levels within a healthy physiological range. At the same time, it considers realistic lifestyle constraints and safety limits for individuals. Simulation results show that the proposed MPC-based strategy improves long-term glycaemic regulation and maintains stability under uncertainties in physiological parameters. The study highlights the potential of control-theoretic approaches as effective tools for preventive healthcare and personalized diabetes management.