Robustness of closed-loop glucose control systems

The main purpose of this work is to provide an extensive, simulation-based comparison of robustness of PID and MPC algorithms in control of blood glucose levels in patients with type 1 diabetes and thus answer the question of their safety. Cohort testing, with 1000 simulated, randomized patients allowed to analyze specific control quality indicators, such as number of hypoglycemic events, and length of hypo- and hyperglycemia periods. Results show that both algorithms provide a reasonable safety level, taking into account natural changes of patients’ physiological parameters. At the same time, we point out drawbacks of each solution, as well as general problems arising in close-loop control of blood glucose level.