dennis bernstein - university of...

1
Questions? Contact: Judi Jones [email protected] Intercalated Injection for Dual Retrospective Adaptive Control of Nonminimum-Phase Systems ABSTRACT: Adaptive control is a promising approach to feedback control of systems with high levels of uncertainty due to unmodeled physics and unpredictable changes. By responding to the characteristics of the actual plant and environment, adaptive control can overcome the performance/robustness tradeoff of robust control. In practice, however, adaptation must be sufficiently fast and accurate despite limited prior modeling information. This talk will focus on recent developments in retrospective cost adaptive control (RCAC) for command following and disturbance rejection. In particular, we will present dual RCAC (DRCAC), which uses concurrent closed- loop identification to estimate the key modeling information needed by RCAC, including nonminimum-phase zeros. This information is used to construct a target model for a closed-loop transfer function that arises from intercalated injection of a virtual exogenous controller perturbation. DRCAC depends on optimization of a biquadratic cost function, whose minimizer provides both the controller update and the required modeling information. This research was performed by current students Yousaf Rahman, Antai Xie, Frant Sobolic, Ankit Goel, Ming Yu, and Ahmad Ansari. BIO: My interests include identification, estimation, and control for aerospace applications. My research has focused on active noise and vibration control, motion control, and spacecraft attitude control. My current interests are in the theory and application of nonlinear system identification, large-scale state estimation for data assimilation, and adaptive control. I am the director of the Noise, Vibration, and Motion Control Laboratory, which includes instrumentation and testbeds for control applications. A 6-degree-of-freedom electric shaker table under all-digital control is used for vibration and motion control applications. Current research is focusing on minimal-modeling adaptive control algorithms for systems with uncertain dynamics and unknown disturbance spectra. I was Editor-in-Chief of the IEEE Control Systems Magazine from August 2003 to December 2011. . Dennis Bernstein University of Michigan Department of Aerospace Engineering Friday, September 16, 2016 3:30 – 4:30 pm • 1500 EECS Sponsored by: Bosch, Ford, and Toyota

Upload: lamkhue

Post on 10-May-2018

215 views

Category:

Documents


1 download

TRANSCRIPT

Questions?Contact:[email protected]

IntercalatedInjectionforDualRetrospectiveAdaptiveControlofNonminimum-PhaseSystems

ABSTRACT: Adaptive control is a promising approach to feedback control of systems with high levelsofuncertainty due to unmodeledphysics andunpredictable changes. By responding to the characteristics of theactual plant and environment, adaptive control can overcome the performance/robustness tradeoff of robustcontrol. In practice, however, adaptation must be sufficiently fast and accurate despite limited prior modelinginformation.Thistalkwillfocusonrecentdevelopmentsinretrospectivecostadaptivecontrol(RCAC)forcommandfollowinganddisturbancerejection.Inparticular,wewillpresentdualRCAC(DRCAC),whichusesconcurrentclosed-loopidentificationtoestimatethekeymodelinginformationneededbyRCAC,includingnonminimum-phasezeros.Thisinformationisusedtoconstructatargetmodelforaclosed-looptransferfunctionthatarisesfromintercalatedinjection of a virtual exogenous controllerperturbation. DRCAC depends on optimization of a biquadratic costfunction,whoseminimizerprovidesboththecontrollerupdateandtherequiredmodelinginformation.

ThisresearchwasperformedbycurrentstudentsYousafRahman,AntaiXie,FrantSobolic,AnkitGoel,MingYu,andAhmadAnsari.

BIO:Myinterestsincludeidentification,estimation,andcontrolforaerospaceapplications.Myresearchhasfocusedonactivenoise and vibration control, motion control, and spacecraft attitude control. My current interests are in the theory andapplicationofnonlinearsystemidentification,large-scalestateestimationfordataassimilation,andadaptivecontrol.Iamthedirector of the Noise, Vibration, and Motion Control Laboratory, which includes instrumentation and testbeds for controlapplications. A 6-degree-of-freedom electric shaker table under all-digital control is used for vibration and motion controlapplications. Current research is focusing on minimal-modeling adaptive control algorithms for systems with uncertaindynamicsandunknowndisturbancespectra.

IwasEditor-in-ChiefoftheIEEEControlSystemsMagazinefromAugust2003toDecember2011..

DennisBernsteinUniversityofMichigan

DepartmentofAerospaceEngineering

Friday,September16,2016

3:30–4:30pm•1500EECS

Sponsoredby:Bosch,Ford,andToyota