Adaptive metaheuristic-optimized PID controllers for closed-loop blood glucose regulation in type-1 diabetes

Type-1 diabetes mellitus requires continuous regulation of blood glucose through appropriate insulin administration, making the development of reliable closed-loop insulin-delivery control strategies an important research problem. This study proposes an intelligent Proportional–Integral–Derivative (PID) control framework for closed-loop blood-glucose regulation based on the Bergman Minimal Model. Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Flower Pollination Algorithm (FPA), and Grey Wolf Optimization (GWO) are employed for PID parameter tuning, with adaptive modifications introduced into their search mechanisms to improve the exploration–exploitation balance and convergence characteristics. A unified optimization framework is established in which each optimization agent represents the PID parameter vector [Kp​,Ki​,Kd​], while its fitness is determined from the closed-loop glucose-regulation performance. The proposed controllers are evaluated through time-domain and frequency-domain analyses and compared with conventional and non-adaptive metaheuristic-tuned PID controllers. Among the investigated approaches, the Adaptive Grey Wolf Optimization-based PID (AGWO-PID) controller demonstrates the most favorable overall performance. For the ISE-based design, AGWO-PID achieves a rise time of 0.296 s and an ISE of 0.235, whereas the ITAE-based design achieves a rise time of 0.315 s, settling time of 0.958 s, and ITAE of 1.066. In addition, descriptive analysis are conducted to account for the stochastic nature of the metaheuristic algorithms, with mean, standard deviation, median, and coefficient of variation used to assess performance consistency. The results demonstrate that adaptive metaheuristic tuning, particularly AGWO-PID, can substantially improve the transient response, tracking performance, and stability of the simulated insulin-delivery system. The findings provide a control-oriented basis for further validation using higher-fidelity physiological models and virtual-patient environments. Ultimately, this research directly supports UN SDG 3 (Good Health and Well-Being) by providing a scalable framework that can contribute to improved healthcare outcomes, enhanced well-being, and broader access to effective health-related solutions.