New therapeutic approaches in diabetes based on novel glucose sensors and programmable insulin pump are nearing clinical introduction, but there are unknowns about the insulin dynamics and quality of control that may be achievable. The overall objective is to develop a set of instructions or control laws for optimal insulin administration based on frequent or continuous glucose determination, such as that provided by novel glucose sensors being developed by us and others. The project will reply on published human data collected by conventional blood glucose sampling, where available, rather than carrying out new experimental studies. Our present methods for blood glucose prediction based on recent blood glucose values will be extended to greater prediction horizons. Systems identification methodology will be used to establish glucose and insulin dynamic models, which will serve as a basis for development and validation of adaptive controllers for glycemic regulation by sensors and insulin pumps. Means will be developed for comparing the degree of control achievable with methods to control based on Hba1c and frequent fingerstick determinations. Specific human data and collection protocols for later application to individual patients will be identified.