Model-based control of nonlinear systems subject to sensor data losses: A chemical process case study

In this work, we consider feedback control of nonlinear uncertain systems subject to sensor data losses. We compare three different potential actions for the controller when feedback is not available: a) setting the input to a fixed value, b) setting the input to the last computed value and c) using the nominal model to predict the value of the state and update the input accordingly. We provide theoretical results that, under certain conditions, motivate the use of feedback based on the nominal model predictions. The results are demonstrated through a chemical process example.