Design of optimized reduced order observer for glucose control with intelligent methods

Many articles associated with glucose-insulin control have been divulged in the last decades, and in these articles frequently supposed that all the system state variables are accessible for feedback. The states like blood glucose and blood insulin are easy to measure, but the measurement of other states such as remote compartment insulin is difficult. This paper proposes an optimized nonlinear Luenberger observer using with the aid of a heuristic algorithm namely Particle Swarm Optimization with Linearly Decreasing Weight (LDW-PSO) for the three-state minimal nonlinear Bergman model. The goal of this optimization is the best recovery of invalid states that are inaccessible or prohibitive to be recovered straightly from the system outputs. The proposed method is a general technique such that for each control input and each disturbance input, one can design an optimal Luenberger observer whereas all unavailable states are appropriately reconstructed. Numerical simulations demonstrate the feasibility of proposed approach.