< Back to previous page

Project

Managing the Complexity and Closed-Loop Performance of Optimal Linear Feedback Controllers for Mechatronic Systems

The complexity of mechatronic systems in industry and society continuously increases because of stricter requirements and an increasing number of functionalities to cope with varying conditions and environments. Associated with that is integrating more and more sensors and actuators, hence their multi-variate character and time-varying dynamics. Designing and tuning controllers for these systems become more and more troublesome, challenging and time-consuming. This increasing control design complexity is partially linked with PID-like control structures, which are inappropriate to cope with the increasing system complexity. Advanced optimal and robust feedback control design techniques can become a worthy alternative to cope with this increasing complexity and become a new standard control design methodology. Their breakthrough in the industry is, however, not yet realized. These optimal robust control techniques are well known in academia. However, industrial control engineers are still reluctant to adopt them because of the lack of burden-less software tools to support the control engineers to design these controllers and their inability to cope with the complexity of the control solution.

The MECO research team of the Department of Mechanical Engineering, KU Leuven, started with the development of the Linear Control Toolbox (LCToolbox) in 2014 [Verbandt, 2019] to facilitate the design of robust optimal controllers and to bring these advanced control design techniques closer to industrial practice. LCToolbox is an open-source Matlab based toolbox that supports a control engineer throughout modeling, identification and advanced feedback controller design. This thesis contributes to this toolbox in the ways described in the following paragraphs.

A novel solver routine is interfaced in the toolbox to design advanced controllers for complex time-varying dynamical systems, specifically linear parameter-varying systems. To further convince the potential of the advanced control design techniques, experimental validations are shown on a lab-scale overhead crane setup with a control design facilitated by the toolbox. Later, the scalability of this solver routine is demonstrated on multi-input multi-output control design using a complex nonlinear system of a 2-DOF manipulator. As a result of interfacing and illustrating the advanced control design techniques, a control engineer can practice unworried and somewhat burden-less control design for complex time-varying dynamical systems.

Furthermore, for complex multi-variate systems, a control engineer may face the challenge of determining the optimal selection/placement of sensors and actuators that manages the controller complexity and achieve desirable closed-loop performance. Although the literature on optimal sensor and actuator selection/placement is established, a systematic methodology to concurrently design advanced controller and attain optimal sensor and actuator selection is lacking. This thesis proposes such a systematic methodology solved using two convex optimization methods that justify the interplay between the controller complexity and the closed-loop performance. The proposed methods are validated on case studies, and the results show that these methods are low-cost and computationally more affordable.  

Besides the selection of sensors and actuators, the complexity of the controller also depends on the complexity of the model and modeled uncertainty, both of which also determine the robustness and performance. To cope with this selection and trade-off, this thesis proposes a systematic methodology to compare models and aid control engineers in selecting a model and uncertainty weight for a robust control design, eventually managing the controller complexity.

Date:22 Nov 2016 →  9 Jul 2021
Keywords:Automation, Optimal and Robust Control
Disciplines:Control systems, robotics and automation, Design theories and methods, Mechatronics and robotics, Computer theory
Project type:PhD project