Exploring the Benefits of Alternative Cost Functions for Nonlinear Model Predictive Control

Quadratic cost functions are commonly used in Nonlinear Model Predictive Control (NMPC) and the algorithms based on them offer excellent control quality for nonlinear processes. The desired control performance can be achieved by properly modifying the values of cost function parameters. The tuning method is often not straightforward and requires a lot of offline simulation experiments, which makes it time-consuming. This work proposes a different, much simpler approach where alternative cost functions are used in place of the classic quadratic one to increase control quality. Simulation experiments were carried out in which the control performance obtained for sixteen NMPC algorithms was evaluated in detail. Each NMPC algorithm had a different cost function and was tested in ideal and noisy environments. The presented benefits of using several alternative cost functions are such as faster control or less wear and tear on the actuator of the controlled system.