Stabilization of the human heartbeat using adaptive controller-based optimized deep policy gradient

Area of Science:

• Biomedical Engineering
• Control Systems
• Computational Neuroscience

Background:

• Cardiac rhythm stabilization is crucial for cardiovascular health and preventing arrhythmias.
• Traditional control methods face challenges due to the complex, dynamic nature of cardiac systems.
• Physiological determinants and external factors introduce significant variability and disturbances.

Purpose of the Study:

• To develop a novel closed-loop control framework for adaptive, real-time cardiac rhythm stabilization.
• To integrate high-order sliding mode control (HO-SMC) with optimized deep policy gradient (ODPG) reinforcement learning.
• To enhance controller robustness against cardiac rhythm uncertainties and disturbances.

Main Methods:

• Combined HO-SMC with ODPG reinforcement learning for adaptive parameter tuning.
• Utilized two neural networks (NNs) for dynamic adjustment of control gains.
• Employed reinforcement learning to evaluate parameter configurations for system stabilization.
• Validated the approach under diverse physiological and pathological conditions.

Main Results:

• Demonstrated superior cardiac rhythm stabilization efficacy compared to conventional controllers.
• Showcased enhanced robustness against uncertainties and disturbances through adaptive parameter tuning.
• Validated the framework's effectiveness across various simulated physiological and pathological scenarios.

Conclusions:

• Pioneered the adaptive synergy of sliding mode control and deep reinforcement learning for cardiac rhythm management.
• The proposed intelligent control system offers a significant advancement in biomedical control.
• This approach enables more effective management of cardiac rhythm irregularities.