Diabetes is a multifaceted chronic disease resulting in the inability of the body to maintain blood glucose concentrations (BGC) within a desired range. Type 1 diabetes (T1D) is caused by the destruction of the beta cells of the pancreas by the immune system, The total loss of insulin production and secretion ability is compensated by exogenous insulin delivery either by injections or infusions by insulin pumps. Treatment by exogenous delivery is much more prevalent than transplantation of a donor pancreas or islets that contain beta cells, which can cure diabetes. Type 2 diabetes (T2D) is also caused by genetic factors and poor food choices and lifestyle. People with T2D have insufficient levels of insulin production or inefficiencies in insulin utilization. T2D can be treated with lifestyle changes and drugs, but insulin may be administered in its advanced stages. The global diabetes prevalence is truly pandemic and has now spread to every corner of the world.
There is no cure for diabetes yet and thus people with diabetes (in particular people with T1D) must actively manage the insulin delivery to their body. Contemporary insulin administration mainly involves daily injections. This was the first solution offered when it was first discovered that diabetes was caused by lack of insulin. Ideally, insulin dosage was estimated and administered depending on the levels of glucose in the blood which was determined by blood/urine analysis and based on meals consumed. Since then, there have been various propositions to automate the insulin administration technique by incorporating various technological developments.
Automation of drug delivery is an active research area full of challenging and complex problems for control engineers and device developers. Inaccuracies in glucose measurements and missing data due to communications problems, delays in tracking BGC with subcutaneous glucose measurement, and delays in diffusion of insulin transport from subcutaneous tissue to the vascular system are compounded with constraints in developing accurate models because of lack of information, uncertainties, nonlinearities and time-varying system parameters. In addition, the response of the body is different for different individuals, and it may change from day-to-day for each person. As such, chronic diseases that necessitate continuous intervention is an important R&D area of drug delivery automation.
Ideally, automating the sequence of activities – measure BGC, estimate meal effect on BGC, determine the dose of insulin to administer and inject/infuse the insulin – would reduce the repetitive activities and ease the burden of people with T1D in regulating their BGC. Technically, automated insulin delivery also called artificial pancreas and has potential to provide a significant benefit.
Two recent special issues in IEEE Control Systems Magazine and Journal of Process Control edited [1.2] by Dr. Ali Cinar, Professor, Departments of Chemical and Biological Engineering and Biomedical Engineering, and Director, Engineering Center for Diabetes Research and Education at Illinois Institute of Technology provided a thorough cross-examination of various technological advents that could be used to automate insulin delivery was undertaken. In papers contributed by leading research groups, various techniques were dissected including technologies tested in clinical trials and simulation studies, highlighting their successes and limitations. Most techniques rely only on automated glucose measurements supplemented by manual entry of information for meals and exercise that cause large perturbations in glucose concentrations.
Professor Cinar is internationally recognized for his research on the next generation of artificial pancreas systems that are fully automated by determining meal information by machine learning and using wearable devices (wristbands) to detect physical activities, psychological stress and sleep characteristics that affect glucose levels [3-5]. His research team is developing machine learning techniques to detect the presence, intensity, characteristics, and duration of these factors and integrating this information in real time with advanced adaptive control techniques to regulate BGC. This approach merges early detection of disturbances that will affect glucose levels (feedforward control with known disturbances) with feedback control based on the recent glucose concentration information. The adaptive personalized model predictive control techniques that they developed uses best estimates of current plasma insulin concentrations in real time and provides accurate insulin infusion dose decisions.
 Cinar A, Artificial Pancreas Systems, IEEE Control Systems Magazine, 2018, 38(1): 26-29.
 Cinar A, Advances in artificial pancreas control systems. Journal of Process Control, volume 81 (2019) page 221–222.Go To Journal of Process Control
 Cinar A, Multivariable Adaptive Artificial Pancreas System in Type 1 Diabetes, Current Diabetes Reports, volume (2017) , page 88
 Hajizadeh I, M Rashid, S Samadi, M Sevil, N Hobbs, R Brandt, A Cinar. Adaptive Personalized Multivariable Artificial Pancreas Using Plasma Insulin Estimates, Journal of Process Control, volume 81 (2019) pages 26-40
 Turksoy K, I Hajizadeh, N Hobbs, J Kilkus, E Littlejohn, S Samadi, J Feng, M Sevil, C Lazaro, J Ritthaler, B Hibner, N Devine, L Quinn, A Cinar. Multivariable Artificial Pancreas for Various Exercise Types and Intensities. Diabetes Technology and Therapeutics, volume 20 (2018) pages 662-671.