In my latest controls homework, we were asked to find a interesting controls problem and assess a case study. It was to learn how multi-variable controls were used to control the system and the techniques that were used. I chose to learn about the Drug Infusion System because understanding how the human body functions is the most relatable subject and it’s interesting to learn how one person sustains himself.
In the first paper that I read, “Issues in the Design of a Multirate Model-Based Controller for a Nonlinear Drug Infusion System”, the Model Predictive Control (MPC) design is used. A model of the process in parallel with the plant to compute the predicted output over a certain number of future sample interval is implemented. The following body parameters were used, Heart rate, Maximum elastance, unstressed venous volume in terms of resistance, systemic resistance, and critical closing pressure. The MPC algorithm consists of two steps. In the prediction step, the model of the process is propagated over the prediction horizon. The unknowns in this propagation are the future values of the manipulated variables. Optimization iterations are required for this part. Nonlinear model is propagated over the prediction horizon prior to optimization, and the linear model is then added on and computed. The multirate nature of the system is then taken care of through correcting disturbances and step by step optimization. A diagram is shown here.
For the second paper that I had read, a multi-variable model reference adaptive control (MRAC) algorithm is developed using a two-input, two-output patient model. The patient hemodynamic model is defined by the linear small-signal first-order transfer function matrix. The control signal (up(t)) that presents the drug infusion rate is formulated as a linear combination of the error feedback (Ke(t) * e) and of the two feed forwards reference model output (Ky(t) * ym(t)) and reference model input (Ku(t) × um(t)). The adaptive control law combines the values of the tracking error “e”, the reference model output “ym” and the reference model input “um” with appropriate adaptive gains (Ky, Ku and Ke). A diagram is shown here.