Incorporating smoothness into weight optimization for H∞ loop-shaping design

Mobolaji Osinuga, Sourav Patra, Alexander Lanzon

    Research output: Chapter in Book/Conference proceedingConference contributionpeer-review

    Abstract

    Smoothness constraints are formulated for weights in H∞ loop-shaping design in order to ensure smooth variations in their magnitude response. Smoothness in the magnitude response of weights prevents the cancelation of lightly damped poles/zeros of the plant when the shaped plant is formed by cascading the nominal plant with the weights. It also allows fitting by low-order transfer functions when the smooth variations are computed point-wise in frequency. Gradients of weights, expressed in dB/decade, are used to formulate the smoothness constraints in LMI form as additional constraints to those on the singular values and condition numbers of weights in [Lanzon, 2005, Weight optimization in H∞ loop-shaping, Automatica, 41(1): 1018-1029]. The resulting solution algorithm maximizes the robust stability margin while simultaneously synthesizing smooth weights and a stabilizing controller subject to the shaped plant lying within a specified region that depicts the closed-loop design requirements. © 2010 IEEE.
    Original languageEnglish
    Title of host publication18th Mediterranean Conference on Control and Automation, MED'10 - Conference Proceedings|Mediterr. Conf. Control Autom., MED - Conf. Proc.
    Pages856-861
    Number of pages5
    DOIs
    Publication statusPublished - 2010
    Event18th Mediterranean Conference on Control and Automation, MED'10 - Marrakech
    Duration: 1 Jul 2010 → …

    Conference

    Conference18th Mediterranean Conference on Control and Automation, MED'10
    CityMarrakech
    Period1/07/10 → …

    Keywords

    • H∞ loop-shaping
    • Robust control
    • Smoothness constraint
    • Weight optimization

    Fingerprint

    Dive into the research topics of 'Incorporating smoothness into weight optimization for H∞ loop-shaping design'. Together they form a unique fingerprint.

    Cite this