Type 1 Diabetes Mellitus (T1DM) is a metabolic disorder in which an individual experiences chronic hyperglycemia (high level of blood glucose), because the pancreatic beta--cells cannot produce or secrete sufficient insulin. Without proper exogenous insulin injection treatment, people with T1DM may experience the complications (vascular damage, kidney failure, or eye damage) related to chronic hyperglycemia, and the complications may eventually result in death. However, the insulin injection treatment also can be harmful, as an over-dose of insulin may result in hypoglycemia (low level of blood glucose) that may result in confusion, coma, or death.

An Artificial Pancreas (AP) is an automated closed-loop system that measures blood glucose concentration and delivers insulin to regulate blood glucose for people with T1DM. Even though there has been significant technological advancement in the AP development, slow absorbing subcutaneous insulin delivery has limited the AP{textquotesingle}s ability to deal with unannounced meal disturbances. As an effort to reduce the actuation delay of the system, rapid acting inhaled and intraperitoneal insulin delivery options have attracted attention as alternative insulin delivery methods for the AP development.

In the work presented in this dissertation, the impact of insulin pharmacokinetics and pharmacodynamics (PK/PD) on performance and robustness of the closed-loop AP was investigated, and two novel APs using enhanced alternative insulin delivery options (inhaled and intraperitoneal routes) were designed and evaluated in simulation and in clinic. Also, a Moving Horizon State Estimator (MHSE) was incorporated into the zone Model Predictive Control (zone-MPC) algorithm to enhance the meal disturbance rejection capacity of the control algorithm.

The theoretical analysis and simulation results regarding the impact of insulin PK/PD on an AP illustrated that the faster acting insulin would result in a faster meal disturbance rejection and tighter glucose regulation. The simulation and clinical evaluations of the semi-automated AP (the AP consists of subcutaneous closed-loop control and inhalation of rapid acting Technosphere RTM Insulin, TI, at meal time) demonstrated that the inhalation of TI at meal time provided the necessary first phase insulin that was not achievable by conventional subcutaneous insulin delivery alone. The semi-automated AP using TI provided faster meal disturbance rejection without imposing extra hypoglycemia risk. The AP using rapid absorbing intraperitoneal insulin also resulted in better glucose regulation (longer time in the clinically accepted safe region, 70-180 mg/dL, and lower postprandial blood glucose peak) compared to the conventional AP using subcutaneous insulin. In addition, the {it in silico} evaluations showed that the MHSE estimated the states more accurately than the Luenberger observer. The zone-MPC using the MHSE responded to the meal earlier and more aggressively compared to the zone-MPC using the Luenberger observer.