A distributed hierarchical iterative learning control framework

Industrial processes, networked systems, and high-precision instruments often involve many subsystems operating in unison on a mutual task. These processes are commonly referred to as systems-of-systems (SOSs) or large-scale systems (LSSs) within engineering fields and are described using multiple-input, multiple-output (MIMO) dynamic system equations. Such systems typically exhibit complex interactions between their constituent subsystems which could be physically collocated or spatially distributed throughout several facilities, e.g., power grids. This makes the control of these distributed processes all the more difficult, especially if communication between the subsystems or their controllers is suboptimal. Many of the operations performed in these settings are done so repeatedly as finite batches of products, e.g., automotive fabrication, or as periodic cycles, e.g., pick-and-place robots on an assembly line. The repetitive nature of these tasks can be leveraged by iterative learning control (ILC) to improve the overall performance of the task. While the most commonly implemented control structures for MIMO systems, learning-based or otherwise, are centralized and decentralized approaches, these architectures have major limitations. Currently available ILC methods and architectures for complex MIMO systems either fail to address the issue of coupling between subsystems adequately, by ignoring them completely or designing overly conservative algorithms for the worst-case scenario or come at a high computational cost and intensive user efforts to implement preprocess decoupling methods for the system models. This dissertation proposes a distributed hierarchical ILC architecture to address these gaps in the literature for a class of complex MIMO systems. The first major contribution of this work is organizing multiple SISO ILC controllers into a multi-level hierarchical structure based on the controlled subsystem dynamics and architecture. Next, we identify a coupling term that is shared among different subsystems of the encompassing ILC controller to mirror the couplings in the controlled MIMO system, thereby connecting them in distributed control manner. We then provide the analytical conditions for the stability and convergence properties of the proposed distributed hierarchical ILC method along with design procedure guidelines to further reduce overall user effort. These proposed methods are then validated on four simulated MIMO systems in comparison with a centralized and decentralized approach.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Mechanical Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Igram, Spencer Scott

Contributors dc:contributor

• Alleyne, Andrew G
• Ferreira, Placid
• Salapaka, Srinivasa
• Stipanovic, Dusan

• optimal control
• distributed control
• iterative learning control
• ILC
• repetitive processes
• linear systems

Statement dc:rights

• Copyright 2022 Spencer Igram

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/116171

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F116171