Switched Controllers in Fully Closed Loop Insulin Delivery Systems: Reducing the Trade-Off Between Prandial Control and Safety.

Abstract

Background: One of the main challenges in control algorithm design for full closed-loop automated insulin delivery systems is the trade-off between the effective compensation of meal-related disturbances and ensuring user safety during the postprandial and fasting periods.

Methods: This paper proposes and evaluates the performance of a switched tuning strategy, a promising but relatively underexplored solution in this domain. This method employs two distinct tunings of a primary control algorithm: an aggressive tuning for meal compensation and a conservative tuning for fasting periods. The analysis considers implementing the switched strategy for three control algorithms: model predictive control and proportional-derivative control, both widely used for glucose regulation, and a linear quadratic Gaussian control, an optimal algorithm previously validated in clinical settings under a switched structure. Additionally, to obtain a more comprehensive understanding of the switched strategy implications, two nonswitched controllers are implemented for each control algorithm: an aggressive and a conservative tuning strategy.

Results: The switched strategy significantly improves the trade-off between meal compensation and safety, increasing the time within the target range of 70-180 [mg/dL] for all three algorithms. For proportional-derivative control, the time in range increases from 69.1% with the conservative tuning and 83.1% with the aggressive to 86.6% with the switched structure. For model predictive control, the improvement is from 73.4% and 74.1% to 85.8%. Last, linear quadratic Gaussian control increases from 65.0% and 70.4% to 85.6%.

Conclusion: The findings suggest that the switched strategy may be a feasible and straightforward approach for enhancing meal compensation without increasing the risk of postprandial hypoglycemia in people with diabetes.