CONICET | Buscador de Institutos y Recursos Humanos

Background and AimsThree short-term artificial pancreas (AP) trials were performed in Argentina from 2016 to 2021 [1]. The last one tested in an outpatient setting both, the Automatic Regulation of Glucose (ARG) algorithm and the InsuMate platform, locally developed. The ARG algorithm is a commuted controller based on a switched LQG algorithm combined with the SAFE layer [2]. MethodIn this work, a new algorithm based on Linear Parameter Varying (LPV) control theory is presented with time-varying capacity to account for insulin sensitivity and other parameter variations during long-term trials. Also, on-line tuning without controller re-design, unannounced meals and physical activity (PA) compensation, and simplified implementation in the InsuMate platform, were sought. Focused on clinical trials, on-line tuning can be made by simply setting two parameters (ϴ1, ϴ2), based on rules that are easy to apply by physicians.ResultsThe proposed AP scheme has been tested with the distribution UVA/Padova simulator under meal and PA disturbances. The results showed that this parameterization achieves similar performance to ideally tuned controllers, and similar or better than our previous ARG algorithm, but with the new added features.ConclusionA new fully parameterized LPV controller has been proposed with the aim of improving online tuning and long-term performance.Simulations show the proposed controller effectiveness.[1] F. Garelli et. al. (2022). First outpatient clinical trial of a full closed-loop artificial pancreas system in South America. Journal of Diabetes Science and Technology. ISSN: 1932-2968. PubMed ID: 35549733.[2] Colmegna, P., Garelli, F., De Battista, H., Sánchez-Peña, R. (2018). Automatic regulatory control in type 1 diabetes without carbohydrate counting. Control Eng Pract(74), 22-32.

enviar mensaje