Non Linear Control of Glycaemia in Type 1 Diabetic Patients Mosè Galluzzo*, Bartolomeo Cosenza Dipartimento di Ingegneria Chimica dei Processi e dei Materiali, Università degli Studi di Palermo Viale delle Scienze, ed. 6, 90128 Palermo, Italy. galluzzo@unipa.it A fuzzy controller for the closed loop control, by insulin infusion of glycaemia in type 1 diabetic patients is proposed. The controller uses type-2 fuzzy sets. The controller was tested in simulation using a complex nonlinear model of the glucose metabolism. Simulation results confirm the effectiveness and the robustness of the type-2 fuzzy logic controller. The design of the controller uses an optimization method based on genetic algorithms. This makes the type-2 fuzzy controller more efficient and faster than a fuzzy controller with type-1 fuzzy sets, allowing a more accurate control of the glucose in the blood. 1. Introduction Diabetes mellitus is one of the most serious and diffuse metabolic diseases and represents a growing problem around the world. The most common forms of diabetes are the so called type 1 and type 2 diabetes. The control problem afforded in this paper is concerned with type-1 diabetes. The main characteristic of type 1 diabetes is that the pancreatic beta-cells fail to produce the insulin needed for the body’s glucidic metabolism requiring that patients inject themselves, several times a day, with external insulin for their survival. The insulin therapy simulates in a dicrete way the pancreas activity. The diabetes has to be kept under control, because abnormal low or high blood glucose levels may lead respectively to cardiovascular problems or to fainting and also to diabetic coma. The development of a completely automatic system, a sort of “artificial pancreas”, that is able to determine and to inject the insulin required by the body’s metabolism on the basis of the measured glucose level in the blood, has been the objective of many researches in the last decades. All the elements that should constitute the closed control loop have been considered: from the sensor of insulin to the infusion system, from the knowledge of all metabolic processes and their mathematical modeling to the more suitable control techniques. The process to be controlled is very complex and in some respects not yet well known. An idea of the problem complexity can be achieved looking at the mathematical models that have been proposed, highly non linear and with parameters that are difficult to determine but in all case strongly affected by uncertainty. In this study the attention is focused on the control technique. Mainly PID and model predictive controllers have been applied to the closed loop control of glucose. Recently a fuzzy controller that makes use of type-2 fuzzy sets has been proposed (Singh et al., 2007). In this paper a type-2 fuzzy logic controller optimized by genetic algorithms is