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

Stabilizing the cardiac rhythm is imperative for preserving cardiovascular health and preventing life-threatening arrhythmias. The stabilization of the heartbeat through traditional control methods presents significant challenges due to the intricate and dynamic characteristics of the cardiac system, which are subject to modulation by a variety of physiological determinants and external perturbations. This study introduces a novel closed-loop control framework combining a high-order sliding mode controller (HO-SMC) with an optimized deep policy gradient (ODPG) reinforcement learning algorithm to achieve adaptive real-time tuning of controller parameters. The integration of HO-SMC with ODPG enables dynamic adjustment of control gains via two neural networks (NNs), enhancing robustness against uncertainties and disturbances inherent in cardiac rhythms. Through reinforcement learning, ODPG evaluates the efficacy of various parameter configurations for stabilizing the system under diverse conditions. As the NNs evolve and fine-tune these parameters, they augment the robustness of the HO-SMC controller, thereby enabling more effective management of uncertainties and disturbances. The proposed approach is validated under various physiological and pathological conditions, demonstrating superior stabilization efficacy compared to conventional controllers. This work pioneers the adaptive synergy of sliding mode control and deep reinforcement learning for cardiac rhythm stabilization, representing a significant advancement in intelligent biomedical control systems.