Insulin Pump Therapy Optimization Based On Run-to-Run Control And Particle Swarm Optimization Algorithm

In recent years, people’s living condition keep improving and living standards are rising with the rapid development of society, while the unhealthy life-style has become the important factor for diabetes mellitus, which have been increasing every day. Because of the miserable complications of diabetes, it has been one of the public health problems, which seriously affect human health of body and mind. Nowdays, many countries have paid much attention on diabetic mellitus. For BG control, various kinds of control algorithms have been successfully implemented, e.g., PID control, model predictive control (MPC), and so on. However, these methods are closed-loop control, which lacks the security, then two improved R2R control algorithm is proposed in below, which is open loop within the batch and closed loop from batch to batch.The first one is an adaptive control strategy combining PSO and R2R control. It has made improvement on the control gain of R2R, using the optimization ability of PSO to obtain the adaptive control gain for R2R control, which depends on experience in the previous literature, so the proposed method can behave more universal. In the simulation, it not only proves the effectiveness of this method, but also shows the excellent control performance when there exist meal variations and subject variability.Another one is high order R2R control strategy. All the reported R2R controllers in glucose regulation are first order. Although the format of first order algorithm is relative concise, it cannot make full use of the historical data, so the control results are not very ideal. In this study, the R2R controller is generalized to high order including the second order and seventh order, and applies in glucose regulation, which there are satisfactory results. In the end, there is a robustness analysis for the high order R2R control algorithm, from the results, it indicates that the high order R2R control can achieve good control effect whether existing meal variations or subject variability.