Learning-based Funnel-MPC for output-constrained nonlinear systems

TL;DR

This paper introduces a learning-based funnel-MPC approach that combines funnel control and model predictive control to improve feasibility, performance, and efficiency for output-constrained nonlinear systems.

Contribution

It presents a novel integration of funnel control with MPC, enhancing feasibility and performance in output-constrained nonlinear control systems.

Findings

Outperforms traditional funnel control in sampling rate and control action

Ensures initial and recursive feasibility in nonlinear MPC

Combines model-free and model-based control for improved learning and performance

Abstract

We exploit an adaptive control technique, namely funnel control, in order to establish both initial and recursive feasibility in Model Predictive Control (MPC) for output-constrained nonlinear systems. Moreover, we show that the resulting feedback controller outperforms the funnel controller both w.r.t. the required sampling rate for a zero-order-hold implementation and required control action. We further propose a combination of funnel control and MPC, exploiting the performance guarantees of the model-free funnel controller during a learning phase and the advantages of the model-based MPC scheme thereafter.