This thesis addresses the problem of model{based control design for blood glucose (BG) concentration of type 1 diabetic patients. Available ontinuous glucose sensors are not reliable enough for adjusting insulin treatment. The problem of initial calibration to relate SC measured glucose to BG is well understood while recalibration techniques are an active area of research. The latter are required because the sensor element gets covered by antibodies thereby continuously altering the transfer characteristics and eventually requiring to replace it every 5{7 days. A novel method for sensor recalibration based on LMIs is presented in this thesis which improves the accuracy of the sensor signicantly when it was tested on real clinical data. Two different variants of the method are presented, one imposing a pure time delay between SC and BG dynamics, the other assuming a first order linear filter. The parameters of those relations are obtained online through convex optimization tools.