Feedback-Driven Self-Optimizing Microservices: Integrating Control Theory, Autonomic Computing, and AI for Adaptive Cloud-Native Systems

Modern cloud-native applications demand high availability, scalability, and resilience, particularly in environments characterized by dynamic workloads, distributed infrastructures, and strict service-level objectives. Microservices architectures, while enabling modularity, independent deployment, and rapid innovation, also introduce significant operational complexity in areas such as service coordination, performance tuning, fault isolation, and resource management. To address these challenges, this paper proposes a feedback-driven control model for self-optimizing microservices that systematically incorporates principles from classical control theory, autonomic computing, and AI-driven observability. By embedding MAPE-K-based feedback loops within individual services and across service ecosystems, the approach enables continuous monitoring of system behavior, intelligent analysis of anomalies and trends, adaptive planning of optimization strategies, and automated execution of corrective or enhancing actions. Furthermore, the integration of machine learning techniques enhances decision-making by enabling predictive scaling, anomaly detection, and policy optimization in real time. Runtime optimization mechanisms such as dynamic load balancing, configuration tuning, and self-healing workflows allow microservices to autonomously adjust performance, resource allocation, and fault recovery without human intervention. This study synthesizes foundational control theory concepts developed since 2000 with recent advancements in self-adaptive and AI-driven systems from 2020 to 2025, ultimately presenting a unified, scalable, and intelligent architecture capable of achieving continuous self-optimization in complex distributed environments while reducing operational overhead and improving system reliability.