## Model Predictive Control

### Background

Model Predictive Control is an advanced control method that been used extensively in the process control industries. The main attraction of MPC is that it enables the plant to operate near the physical and/or safety limits, which can increase productivity and reduce product quality variation. This ability stems from the fact that MPC calculates the next control action online via the solution of an optimisation problem, and system constraints and physical limits are naturally included at this stage as inequality constraints on the variables of the optimisation problem. This optimal control action will be applied at the next sample instant, where new information is also measured from the plant. Based on this new information the entire process repeats and a new optimisation problem is solved to provide the next control action. And so on.

As one may expect, solving such an optimisation problem in the time available (between sample points) can be a very challenging task. Fortunately for the process control industry, the associated plant dynamics are typically slow so that the sample interval is in the order of seconds, minutes, hours or even days. This allows plenty of time to solve the MPC optimisation problem and to determine the next control move.

However, application of MPC to systems that exhibit fast dynamic behaviour becomes even more challenging because the time constraint on solving the optimsation problem can be in the order of milli or even micro seconds. This class of problems requires a new approach, where the optimisation solver is geared toward speed as well as accuracy.

This project is addressing the latter class of fast systems with the development of high-speed MPC solutions. A selection of pilot studies can be found here