Robust approximation to adaptive control by use of representative parameter sets with particular reference to type 1 diabetes

Control of many biological processes can involve significant difficulties. Even when a suitable control algorithm has been devised for a model of the system, state and parameter estimation can be difficult because of time-lags in measurement availability due to inherent delays in chemical or other analytical techniques. The problem is further compounded when adaptive control is required due to the likelihood of system parameter changes over time. An example is the case of control of plasma glucose levels in Type 1 (insulin dependent) diabetic subjects, a nonlinear problem where glucose/insulin model parameter estimation requires clinical tests which take too long to be useful in the dynamic situation. The parameters of the glucose/insulin dynamics model may also vary over time for an individual due to factors such as sleep, exercise and level of health. We describe a straight-forward approach which may be useful in such situations.