Optimal Modular Control of Overactuated Systems - Theory and Applications.

Decentralized Control;Robust Control;Model Predictive Control;Automotive Control;Electrical Engineering;Engineering;Electrical Engineering: Systems

Vermillion, Christopher R.Stefanopoulou, Anna G. ;

University of Michigan

关键词: Decentralized Control;    Robust Control;    Model Predictive Control;    Automotive Control;    Electrical Engineering;    Engineering;    Electrical Engineering: Systems;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/62379/cvermill_1.pdf?sequence=1&isAllowed=y

瑞士|英语

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

【 摘 要 】

This dissertation develops design and analysis tools for the modular control of overactuated systems.Modular control has enabled control system designers to decompose the control design into simpler subtasks that can be addressed in parallel.This dissertation focuses on a particular inner loop/outer loop modular control strategy that has been popularized for the control of overactuated systems.In spite of its appeal, this control strategy brings about several stability and performance-related challenges.In particular, existing techniques for guaranteeing overall stability of the system are either very conservative or require a very large degree of time-scale separation between the inner and outer loops.In addition, the current literature does not provide tools that can be used to address the performance tradeoff between the modular control design and its centralized counterpart.This dissertation addresses these challenges by developing a new modular design framework, which is based on an inner loop emph{reference model} that serves as a performance specification for the inner closed loop.This approach allows the designer to address overall system stability and performance without relying on significant time scale separation between the inner and outer loops.The use of the reference model also facilitates the derivation of new analytical results that quantify the performance gap between the modular and centralized strategies.On top of the reference model based approach, the dissertation proposes a novel mechanism, referred to as modular control error compensation (MCEC), for recovering performance when the inner closed loop does not match the reference model.Finally, the dissertation explores the use of model predictive control allocation (MPCA) as an optimization-based inner loop control strategy.As a case study, the reference model based design and accompanying control strategies (MCEC and MPCA) are experimentally validated on a thermal management system that is used for engine modeling in an automotive test cell.Extensive modeling and dynamic analysis results are first presented, exhibiting a high fidelity dynamic model that is well suited for control design.The dissertation proceeds to use this dynamic model for control design, presenting experimental results that show the benefit of the proposed control methodologies.