Control strategies and AI-based techniques for blood glucose regulation in artificial pancreas system development: A systematic literature review

Diabetes mellitus, a chronic metabolic disorder, affects millions worldwide, making effective regulation of blood glucose levels (BGL) vital to prevent complications such as hyperglycemia and hypoglycemia. The emergence of Artificial Pancreas Systems (APS), which combine continuous glucose monitoring (CGM) with automated insulin delivery (AID), has revolutionized diabetes management. However, the effectiveness of APS largely depends on the choice and implementation of control strategies. While several reviews have examined blood glucose regulation, many adopted non-systematic approaches and concentrated mainly on linear controllers like Proportional-Integral-Derivative (PID) methods, with limited focus on nonlinear and artificial intelligence (AI)-driven techniques. This study systematically reviews recent advancements in APS control strategies following the PRISMA guidelines. A comprehensive search of major databases, including ScienceDirect, SpringerNature, IEEE, MDPI, SAGE, and Wiley, identified 881 articles, of which 51 met the inclusion criteria for detailed analysis. The findings reveal substantial progress in both linear and nonlinear control methods, with AI-based techniques demonstrating significant potential for personalized and adaptive glycemic management. Nonetheless, persistent challenges remain regarding computational efficiency, real-time performance, and clinical validation. The study highlights existing research gaps and proposes a roadmap for future investigations to enhance APS design and functionality. By advancing robust, intelligent, and clinically viable control frameworks, this review contributes to the development of next-generation APS solutions aimed at improving the quality of life for individuals living with diabetes. • Systematic PRISMA-based review of linear and nonlinear Artificial Pancreas Systems (APS) control methods for Type 1 diabetes management • Comparative analysis of PID, model-based, and Artificial Intelligence (AI) driven controllers for personalized glucose regulation • Evaluation of computational complexity and real-time implementation challenges in APS • Identification of gaps in validation, robustness, and real-world clinical deployment • Future research directions for adaptive APS design and improved clinical outcomes