A simulation study on adaptive control design performance for T1DM via individualized model

In Type 1 Diabetes Mellitus (T1DM), external administration of insulin is needed due to absolute deficiency of insulin production in pancreas. Model reference adaptive control (MRAC) is considered because of possibility of time varying parameters of insulin-glucose system. T1DM simulator is individualized by parameter identification of nonlinear model on the basis of continuous glucose monitoring (CGM) data. MRAC is then designed using linear approximation of nonlinear model. Due to disturbance inputs, which represent carbohydrate intake, the feedforward disturbance rejection and modifications for robustness of adaptive law are added. Performance evaulation is done by comparing individualized model simulation output with CGM data, which represent manual administration of insulin. Furthermore the robustness of control design is evaulated, using generated population of virtual T1DM patients via reasonable random changes in model parameters, by means of error grid analysis.