Search NASA⌕ Search

Engineering topics

Byrnes, C. I.

Publications and source records attributed to Byrnes, C. I..

At least 19 records
All sourcesNASA NTRSDOE OSTI
NASA NTRS1992

Singularity perturbed zero dynamics of nonlinear systems

Stability properties of zero dynamics are among the crucial input-output properties of both linear and nonlinear systems. Unstable, or 'nonminimum phase', zero dynamics are a major obstacle to input-output linearization and high-gain designs. An analysis of the effects of regular perturbations in system equations on zero dynamics shows that whenever a perturbation decreases the system's relative degree, it manifests itself as a singular perturbation of zero dynamics. Conditions are given under which the zero dynamics evolve in two timescales characteristic of a standard singular perturbation form that allows a separate analysis of slow and fast parts of the zero dynamics.

Isidori, A.↗
NASA NTRS1992

Decentralized-feedback pole placement of linear systems

A projectile product spaces model is used to analyze decentralized systems. The degree of the pole placement map is computed. The conditions under which the degree is odd are also given. Twin lift systems are studied. It is proved that the poles of a twin lift system can be assigned to any values by local static and local dynamic feedback laws if and only if the system is jointly controllable.

Wang, X.↗
NASA NTRS1984

Output feedback and generic stabilizability

Attention is given to questions of pole placement and stabilization for generic linear systems with prescribed state, input and output dimension for the case where the controller must be implemented by linear memoryless output feedback. A criterion is presented in terms of a special pole placement property for generic stabilizability, and this criterion is used to describe constraints on the dimension which are consistent with generic stabilizability. Also treated are the rationality and solvability by radicals of stabilizing or pole positioning gains, and the way in which decision algebra can theoretically handle existence questions for generic systems is described.

Byrnes, C. I.↗
NASA NTRS1983

Simultaneous stabilization and simultaneous pole-placement by nonswitching dynamic compensation

The 'simultaneous stabilization problem' is defined and theorems are proposed for its solution. The problem consists in answering the question: given an r-tuple G sub 1(s), G sub r(s) of p x m proper transfer functions, does there exist a compensator K(s) such that the closed loop systems G sub 1(s) (I+K(s)G sub 1(s)) (-1), G sub r(s) (I+K(s) G sub r(s)) (-1) are (internally) stable. This question arises in reliability theory, where G sub 2(s), G sub r(s) represents a plant G sub 1(s) operating in various modes of failure and K(s) is a nonswitching stabilizing compensator. It is important in the stability analysis and design of a plant which can be switched into various operating modes. The simultaneous stabilization problem can also apply to the stabilization of a nonlinear system which is linearized at several equilibria. Conditions are defined for pole placement and the generalized Sylvestor matrix is discussed. Previously announced in STAR as N82-31031

Ghosh, B. K.↗
NASA NTRS1983

On the stabilizability of multivariable systems by minimum order compensation

In this paper, a derivation is provided of the necessary condition, mp equal to or greater than n, for stabilizability by constant gain feedback of the generic degree n, p x m system. This follows from another of the main results, which asserts that generic stabilizability is equivalent to generic solvability of a deadbeat control problem, provided mp equal to or less than n. Taken together, these conclusions make it possible to make some sharp statements concerning minimum order stabilization. The techniques are primarily drawn from decision algebra and classical algebraic geometry and have additional consequences for problems of stabilizability and pole-assignability. Among these are the decidability (by a Sturm test) of the equivalence of generic pole-assignability and generic stabilizability, the semi-algebraic nature of the minimum order, q, of a stabilizing compensator, and the nonexistence of formulae involving rational operations and extraction of square roots for pole-assigning gains when they exist, answering in the negative a question raised by Anderson, Bose, and Jury (1975).

Byrnes, C. I.↗
NASA NTRS1982

Simultaneous stabilization and simultaneous pole placement by nonswitching dynamic compensation

The 'simultaneous stabilization problem' is defined and theorems are proposed for its solution. The problem consists in answering the question: given an r-tuple G sub 1(s), G sub r(s) of p x m proper transfer functions, does there exist a compensator K(s) such that the closed loop systems G sub 1(s) (I+K(s)G sub 1(s)) (-1), G sub r(s) (I+K(s) G sub r(s)) (-1) are (internally) stable. This question arises in reliability theory, where G sub 2(s), G sub r(s) represents a plant G sub 1(s) operating in various modes of failure and K(s) is a nonswitching stabilizing compensator. It is important in the stability analysis and design of a plant which can be switched into various operating modes. The simultaneous stabilization problem can also apply to the stabilization of a nonlinear system which is linearized at several equilibria. Conditions are defined for pole placement and the generalized Sylvestor matrix is discussed.

Ghosh, B. K.↗
NASA NTRS1982

The McMillan and Newton polygons of a feedback system and the construction of root loci

The local behaviour of root loci around zeros and poles is investigated. This is done by relating the Newton diagrams which arise in the local analysis to the McMillan structure of the open-loop system, by means of what we shall call the McMillan polygon. This geometric construct serves to clarify the precise relationship between the McMillan structure, the principal structure, and the branching patterns of the root loci. In addition, several rules are obtained which are useful in the construction of the root loci of multivariable control systems.

Byrnes, C. I.↗
NASA NTRS1982

Pole placement by static and dynamic output feedback

This paper gives new results concerning pole-assignability by static and dynamic output feedback, based on the interpretation of transfer functions, feedback laws, poles and zeroes in terms of the incidence geometry of m-planes and p-planes in (m+p)-space. As an illustration of the most basic ideas, a short proof of the Brasch-Pearson theorem is given. A more careful analysis of this proof yields a significant extension of this theorem, which is considerably sharpened in the case of pole-assignment by constant gain output feedback. As a final application, a root-locus design technique for non-square systems is introduced: zero placement by pre- or post-compensation. This zero placement problem is then analyzed by methods similar to those developed for pole placement by output feedback.

Byrnes, C. I.↗
NASA NTRS1981

Multivariable Nyquist criteria, root loci, and pole placement - A geometric viewpoint

A description is given of what are considered to be the natural multivariable analogs of concepts from classical control theory. A satisfactory generalization of the Nyquist criterion is described, and a clear analog of the asymptotic properties of the root locus is obtained in the 'multiparameter' case. An example is given, however, which illustrates the quite surprising fact that the root locus map is not always continuous at infinite gains. It is noted that this calls for a new ingredient, a compactification of the space of gains, and that what is the most interesting new feature in this circle of ideas comes in the area of pole placement.

Brockett, R. W.↗
NASA NTRS1980

Realization theory and quadratic optimal controllers for systems defined over Banach and Frechet algebras

It is noted that recent work by Kamen (1979) on the stability of half-plane digital filters shows that the problem of the existence of a feedback law also arises for other Banach algebras in applications. This situation calls for a realization theory and stabilizability criteria for systems defined over Banach for Frechet algebra A. Such a theory is developed here, with special emphasis placed on the construction of finitely generated realizations, the existence of coprime factorizations for T(s) defined over A, and the solvability of the quadratic optimal control problem and the associated algebraic Riccati equation over A.

Byrnes, C. I.↗
NASA NTRS1980

Hamiltonian indices and rational spectral densities

Several (global) topological properties of various spaces of linear systems, particularly symmetric, lossless, and Hamiltonian systems, and multivariable spectral densities of fixed McMillan degree are announced. The study is motivated by a result asserting that on a connected but not simply connected manifold, it is not possible to find a vector field having a sink as its only critical point. In the scalar case, this is illustrated by showing that only on the space of McMillan degree = /Cauchy index/ = n, scalar transfer functions can one define a globally convergent vector field. This result holds both in discrete-time and for the nonautonomous case. With these motivations in mind, theorems of Bochner and Fogarty are used in showing that spaces of transfer functions defined by symmetry conditions are, in fact, smooth algebraic manifolds.

Byrnes, C. I.↗
NASA NTRS1980

Recent results on output feedback problems

Given a real linear system sigma = (A, B, C) with m inputs, p outputs and degree n, the problem of generic pole placement by output feedback is studied, which is to compute the constant C(m,p) such that the inequality C(m,p) not less than n is necessary and sufficient for generically positioning the poles of the generic linear system by constant output feedback. A constant C prime (m,p) is determined, which gives a sufficient condition for generic pole placement and which, to the best of the author's knowledge, is at least as good an estimate of C(m,p) as any in the literature. Some results on the construction of solutions in case mp = n are announced, based on the degree formula of Brockett and Byrnes and the Galois theory. In particular, a question raised by Anderson, Bose, and Jury, on the existence of a rational procedure for computing the feedback law from the desired characteristic polynomial is answered.

Byrnes, C. I.↗
NASA NTRS1979

On certain families of rational functions arising in dynamics

It is noted that linear systems, depending on parameters, can occur in diverse situations including families of rational solutions to the Korteweg-de Vries equation or to the finite Toda lattice. The inverse scattering method used by Moser (1975) to obtain canonical coordinates for the finite homogeneous Toda lattice can be used for the synthesis of RC networks. It is concluded that the multivariable RC setting is ideal for the analysis of the periodic Toda lattice.

Byrnes, C. I.↗
NASA NTRS1979

Feedback invariants for linear systems defined over rings

A coefficient-assignability theorem is presented for systems defined over a commutative ring with 1. The conditions examined include all the general results on coefficient-assignability available in the literature. It is concluded that the proposed techniques are not effective for the case of the weaker property of pole placement when it is assumed that the commutative ring is a PID. Morse's well-known theorem seems to be the best result available for this situation.

Byrnes, C. I.↗