Integrating Multiple Controllers to Balance Competing Performance Objectives: Application to Blood Glucose Management

This paper develops a control integration methodology that can be used to balance competing performance objectives. The target application for the method is blood glucose management for diabetic patients. The true system model is uncertain and changes over time. Multiple controllers have been designed with competing performance objectives in mind. In the proposed method, on-line predictors are used to estimate the future state of the system based on the current state and the different control recommendations. The system then switches between the different controllers based on the predicted outcomes from their control recommendations. Simulation results demonstrate that the integrated approach based on predicted outcomes performs better across a large range of disturbances than any of the individual controllers.