Agent-enabled adaptive and fault-tolerant control for healthcare electromechanical systems under time-varying physical constraints

Abstract

Electromechanical systems in healthcare cyber-physical systems (CPS), such as remote surgical robots, require adaptive and fault-tolerant control mechanisms to maintain stability under time-varying physical constraints. However, conventional control strategies often struggle with semantic gaps between multi-source measurements and low-level control actions, failing to adapt efficiently to sudden disturbances or dynamic conditions. To address these challenges, this paper proposes an agent-enabled, intent-driven adaptive control framework. By seamlessly integrating multi-source sensing with closed-loop control execution, we design a comprehensive fourlayer architecture featuring semantic-aware intent parsing, distributed decision-making, and dynamic resource orchestration. Furthermore, to achieve a reliable mapping from high-level operational intents to fine-grained, constraint-aware control execution, a deep reinforcement learning mechanism based on proximal policy optimization is incorporated. Extensive simulation results demonstrate that the proposed framework significantly reduces the end-to-end response delay and substantially shortens the execution adaptation time. Notably, it achieves a high constraint satisfaction rate under dynamic disturbance scenarios, thereby effectively guaranteeing fault tolerance, stability, and system robustness for complex electromechanical applications in smart healthcare.