Multi-Objective Iterative Learning Control: An Advanced ILC Approach for Application Diversity.

Iterative Learning Control;Mechanical Engineering;Engineering;Mechanical Engineering

Lim, IngyuErsal, Tulga ;

University of Michigan

关键词: Iterative Learning Control;    Mechanical Engineering;    Engineering;    Mechanical Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/120875/ingyulim_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

While ILC has been applied to repetitive applications in manufacturing, chemical processing, and robotics, several key assumptions limit the extension of ILC to various applications. Conventional ILC focuses on improving the performance of a single metric, such as tracking performance through iterative updates of the time domain control input. The application range is limited to systems that satisfy the assumption of iteration invariance of the plant, reference signal, initial conditions, and disturbances.We aim to relax this assumption to gain significant advantages. More specifically we focus on relaxing the strict reference tracking requirement to address multiple performance metrics and define the stability bounds across temporal and spatial domains. The aim of this research is expanding the application space of ILC towards non-traditional applications.Chapter III presents an initial framework to provide the foundation for the multi-objective ILC. This framework is validated by simulation and experimental tests with a wheeled mobile robot.Chapter IV extends the initial framework from the temporal domain to the spatial domain. The initial framework is generalized to address four classifications of performance objectives. Stability and performance analysis for each classification is provided. Simulation results on a high-resolution additive manufacturing system validate the extended framework. For the generalized framework, we present a distributed approach in which additional objectives are considered separately. Chapter V evaluates the difference between this distributed approach, and a centralized approach in which the objectives are combined into a single matrix depending on the classification. Chapter VI extends the multi-objective ILC to incorporate a region-based tracking problem in which reference uncertainty is addressed through the development of a bounded region. A multi-objective region-to-region ILC is developed and validated by a simulation of a surveillance problem with an UAV and multiple unattended ground sensors. Comparisons with point-to-point ILC, region-to-region ILC, and multi-objective region-based ILC demonstrate the performance flexibility that can be achieved when leveraging the regions. This dissertation provides new approaches for relaxing the classical assumption of iteration invariant reference tracking. New stability and convergence analysis is provided, resulting in a design methodology for multi-objective ILC. These approaches are validated by simulation and experimental results.