• DocumentCode
    358763
  • Title

    Identifiability issues in general interconnected systems

  • Author

    Claassen, Mareike ; Wemhoff, Eric ; Packard, Andrew ; Poolla, Kameshwar

  • Author_Institution
    Dept. of Mech. Eng., California Univ., Berkeley, CA, USA
  • Volume
    1
  • Issue
    6
  • fYear
    2000
  • fDate
    36770
  • Firstpage
    230
  • Abstract
    We investigate the problem of identifiability in general interconnected systems. We begin with an arbitrary, structured interconnection of two types of blocks: static nonlinear maps and linear time invariant systems. Indeed, any nonlinear time-invariant dynamical system may be represented as such an interconnection. The blocks may be completely specified or may be unknown and require identification. The input-output dimensions of the unknown blocks are fixed. No other a priori information about them is available. In particular, the static nonlinear maps requiring identification are not naturally parameterized via basis functions or other expansions. The state dimensions of the LTI blocks that require identification are unavailable. Thus, we are interested in non-parametric identification problems. The particular focus of the paper is the issue of “identifiability”. Loosely speaking, a general structured interconnection is identifiable if it is possible to determine the unknown blocks uniquely using input-output experiments alone. As is well known, identifiability concepts are of fundamental importance in system identification. We begin an investigation of identifiability in this context. We offer a characterization of identifiability under certain restrictive assumptions. The computational aspects of this characterization remain to be explored
  • Keywords
    discrete time systems; identification; interconnected systems; matrix algebra; nonlinear systems; general interconnected systems; identifiability issues; input-output dimensions; linear time invariant systems; static nonlinear maps; Convergence; Interconnected systems; Kernel; Linearity; Mechanical engineering; Neural networks; Nonlinear systems; Reduced order systems; System identification; Time invariant systems;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    American Control Conference, 2000. Proceedings of the 2000
  • Conference_Location
    Chicago, IL
  • ISSN
    0743-1619
  • Print_ISBN
    0-7803-5519-9
  • Type

    conf

  • DOI
    10.1109/ACC.2000.878842
  • Filename
    878842