Artificial Pancreas System With Unannounced Meals Based on a Disturbance Observer and Feedforward Compensation

Abstract

This paper is focused on closed-loop control of postprandial glucose levels of patients with type 1 diabetes mellitus after unannounced meals, still a major challenge towards a fully autonomous artificial pancreas. The main limitations are the delays introduced by the subcutaneous insulin pharmacokinetics and the glucose sensor, which typically lead to insulin over-delivery. Current solutions reported in the literature typically resort to meal announcement, which requires the patient intervention. In this paper, a disturbance observer is used to estimate the effect of unannounced meals and the insulin pharmacokinetics is taken into account by means of a feedforward compensator. The proposed strategy is validated in silico with the UVa/Padova metabolic simulator. It is demonstrated how the disturbance observer successfully estimates and counteracts not only the effect of meals but also sudden drops in the glucose levels that may lead to hypoglycemia. For unannounced meals, results show a median time-in-range of 80% in a 30-day scenario with high carbohydrate content and large intra-subject variability. Optionally, users may decide to announce meals. In this case, considering severe bolus mismatch due to carbohydrate counting errors, the median time-in-range is increased up to 88%. In every case, hypoglycemia is avoided.

Fault Tolerant Strategies for Automated Insulin Delivery Considering the Human Component: Current and Future Perspectives

“…(1) estimating insulin on board (IOB) or plasma insulin concentrations, 39,40 (2) glucose rate of appearance after meal estimations, [41][42][43] (3) exercise detection, [44][45][46] and (4) general state estimation for prediction and control. [47][48][49][50][51][52]…”

“…Several algorithms detect when a meal has been consumed, largely based on glucose rate-of-change greater than a threshold value, and therefore become more aggressive with insulin delivery. 52,[60][61][62][63] Indeed, knowledge of typical daily eating patterns can be used to anticipate when meals are likely to be consumed and include that probability in future predictions of BG levels. 50,51,[64][65][66] While these anticipatory algorithms lead to good CL performance, providing a premeal bolus yields a significant improvement in meal disturbance rejection.…”

A Hybrid Automata Approach for Monitoring the Patient in the Loop in Artificial Pancreas Systems

“…Developed strategies to detect specific patient behaviors exist, however, they were not developed for safety but as ad-hoc approaches for control. Most of the approaches involving the patient estimate the glucose rate of appearance to allow AP to operate without meal announcement [18][19][20][21][22][23]. Other strategies try to detect when patients exercise and modify or adapt their control strategy accordingly [24,25].…”

Novel Strategies to design Controllers and State Predictors based on Disturbance Observers

“…In addition, their application to other elds, such as: economics (Bernhard 2003;Hansen et al 2008), game-theory (Ba³ar et al 1998) or biotechnology (Sala-Mira et al 2019;Sanz et al 2020); is also being explored.…”

“…We started by looking at the problem of closed-loop glucose-control for people with type-1 diabetes; e.g. (Tecnodiabetes ai2, UPV ; Sanz et al 2020). The control problem consist in the following: measures of the glucose concentration in blood are taken and it is needed to develop a controller that based on these measurements automatically injects insulin to the person if needed.…”