Gaussian Process Nonlinear Model Predictive Control for Online Partially Observable Systems: An Application to Froth Flotation

Abstract

A nonlinear model predictive control framework that uses Gaussian process state-space models, trained on data from a physics-based flotation model, when only some process states can be measured online. Because the Gaussian processes predict both mean and variance, the controller accounts for uncertainty and disturbances. It reduced valuable mineral losses to tailings by 20% compared with traditional control, tracked setpoints within 5% and runs within real-time limits.

Dr Paulina Quintanilla

Assistant Professor in Process Systems Engineering

Assistant Professor in Process Systems Engineering at UCL, leading the LOOPS research group. Research interests include machine learning, optimisation and control for processes and physical systems.