Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimized Pseudo-Linearization-Based Model Predictive Controller Design: Direct Data-Driven Approach

Mikiya Sekine, Satoshi Tsuruhara, Kazuhisa Ito

TL;DR

The paper addresses the challenge of selecting an appropriate reference model in FRIT for nonlinear control. It introduces a pseudo-linearization (PL) approach that pairs a PL-based inner loop with an outer-loop MPC, enabling data-driven tuning while explicitly handling input constraints. The method estimates internal plant inputs via a simple PL state-space model and optimizes the MPC cost to reflect designer intent, demonstrated through Hammerstein and Bouc–Wen simulations and experimental tests on tap-water–driven artificial muscles. Results show direct tracking of the reference signal r(k) with reduced dependence on a fixed reference model, improved RMSE in nonlinear scenarios, and practical applicability to systems with hysteresis. The work also outlines future directions for robustness through adaptive FRIT and improved inner-loop nonlinear handling.

Abstract

To reduce the typical time-consuming routines of plant modeling for model-based controller designs, the fictitious reference iterative tuning (FRIT) has been proposed and has proven to be effective in many applications. However, it is generally difficult to select a reference model properly without information on the plant, which significantly affects the control performance and sometimes leads to considerable performance degradation. To address this problem, we propose a pseudo-linearization (PL) method using FRIT and design a new controller for nonlinear systems that combines data-driven and model-based control. This design considers the input constraints using model predictive control. The effectiveness of the proposed method was evaluated according to several practical references using numerical simulations for nonlinear classes and experiments involving artificial muscles with hysteresis characteristics.

Optimized Pseudo-Linearization-Based Model Predictive Controller Design: Direct Data-Driven Approach

TL;DR

The paper addresses the challenge of selecting an appropriate reference model in FRIT for nonlinear control. It introduces a pseudo-linearization (PL) approach that pairs a PL-based inner loop with an outer-loop MPC, enabling data-driven tuning while explicitly handling input constraints. The method estimates internal plant inputs via a simple PL state-space model and optimizes the MPC cost to reflect designer intent, demonstrated through Hammerstein and Bouc–Wen simulations and experimental tests on tap-water–driven artificial muscles. Results show direct tracking of the reference signal r(k) with reduced dependence on a fixed reference model, improved RMSE in nonlinear scenarios, and practical applicability to systems with hysteresis. The work also outlines future directions for robustness through adaptive FRIT and improved inner-loop nonlinear handling.

Abstract

To reduce the typical time-consuming routines of plant modeling for model-based controller designs, the fictitious reference iterative tuning (FRIT) has been proposed and has proven to be effective in many applications. However, it is generally difficult to select a reference model properly without information on the plant, which significantly affects the control performance and sometimes leads to considerable performance degradation. To address this problem, we propose a pseudo-linearization (PL) method using FRIT and design a new controller for nonlinear systems that combines data-driven and model-based control. This design considers the input constraints using model predictive control. The effectiveness of the proposed method was evaluated according to several practical references using numerical simulations for nonlinear classes and experiments involving artificial muscles with hysteresis characteristics.
Paper Structure (15 sections, 23 equations, 19 figures, 4 tables, 1 algorithm)

This paper contains 15 sections, 23 equations, 19 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Data--driven Control
  3. Model-based Control System Using PL
  4. PL using E--FRIT
  5. Estimation of Internal Variables
  6. MPC Design with PL Model
  7. Simulation Study
  8. Simulation: Hammerstein Model
  9. Asymmetric Bouc--Wen Model
  10. Experimental Setup and Simulation Conditions
  11. Simulation: Asymmetric Bouc--Wen Model
  12. Case Study: Tap--water--driven Artificial Muscle Control System
  13. Experimental Conditions
  14. Experimental Results and Discussion
  15. Conclusion

Figures (19)

  • Figure 1: Block diagram for FRIT method.
  • Figure 2: Block diagram for model--based controller using PL with E--FRIT
  • Figure 3: Block diagram for adaptive E--FRIT--based MPC
  • Figure 4: Comparison of tracking performance between the proposed and conventional methods in simulations of the Hammerstein model
  • Figure 5: Control input $u(k)$ to the plant in simulations of the Hammerstein model
  • ...and 14 more figures

Theorems & Definitions (3)

  • remark 1
  • remark 2
  • remark 3