Concurrent Learning Adaptive Model Predictive Control with Pseudospectral Implementation

TL;DR

This paper introduces a novel control architecture combining concurrent learning-based adaptive control with model predictive control, utilizing pseudospectral methods for state derivative approximation, demonstrated through theoretical analysis and simulations.

Contribution

It develops a unified control framework that integrates adaptive learning directly into MPC without switching phases, employing pseudospectral techniques for improved numerical accuracy.

Findings

Effective compensation for model uncertainties demonstrated

Unified control and learning process validated theoretically

Numerical simulations confirm improved control performance

Abstract

This paper presents a control architecture in which a direct adaptive control technique is used within the model predictive control framework, using the concurrent learning based approach, to compensate for model uncertainties. At each time step, the control sequences and the parameter estimates are both used as the optimization arguments, thereby undermining the need for switching between the learning phase and the control phase, as is the case with hybrid-direct-indirect control architectures. The state derivatives are approximated using pseudospectral methods, which are vastly used for numerical optimal control problems. Theoretical results and numerical simulation examples are used to establish the effectiveness of the architecture.