Browsing by Author "Pearson, J. B."

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    A dead beat controller and least squares parameter estimator for a three joint robot arm
    (1985) Barry, Patricia A.; Pearson, J. B.; Kohn, Wolf; Briggs, Faye A.
    A servo-controller is developed for a three joint manipulator. The servo-controller consists of a dead beat controller and an on-line parameter estimator. The dead beat controller is implemented in four steps. A piecewise linear model is calculated on line. Simultaneous diagonalization of the stiffness and inertia matrices is performed after linearization. A control vector for the linearized, diagonalized system is generated at a faster rate. The control vector is transformed from the diagonalized mode to the real mode via multiplication by a transformation matrix. The parameter estimator is a least' squares estimator designed to estimate the stiffness parameters in the diagonalized mode. The parameter estimator iteratively estimates the eigenvectors and eigenvalues associated with the stiffness parameters. The parameter estimator converges too slowly to meet the design requirements. The controller, however, provides an acceptable tracking error* and is robust enough to use without the parameter estimator.
  • Thumbnail Image
    Item
    Computational methods in the design of linear control systems
    (1980) Kontos, Athanasios V.; Pearson, J. B.; Parks, Thomas W.; Figueiredo, Rui J. P. de
    This thesis considers the problem of computing controllers for multivariable systems. System representation is in terms of polynomial matrices and two algorithms are presented which are shown to be useful in the design of controllers for such systems. These algorithms are: i) Factorization of a polynomial matrix, and ii) Computation of a unimodular matrix U satisfying the relation [A B]U = I 1, where A and B are left coprime polynomial matrices. These algorithms do not involve numerically unsatisfactory Euclidean type operations. It is shown that the two algorithms can be used to compute solutions to the system stabilization problem and to the model matching problem. The Regulator Problem with Internal Stability (RPIS) is also discussed, and under certain assumptions it is shown how solutions can be computed.
  • Thumbnail Image
    Item
    The general linear multivariable regulator problem
    (1975) Cheng, Lin-Fu; Pearson, J. B.
    This thesis considers the problem of zeroing the output z(t) of a general linear, fixed-parameter, multivariable system as described by x(t) * Ax(t) + Bu(t) z(t) = C,jx(t) + D?X> and(3) any controllable, observable mode of the closed-loop system is stabilized. Vector space manipulations are adopted throughout the thesis. First, an algebraic formulation corresponding to (1) is derived and the problem in its most primitive form is solved. Since output (y(t)) feedback is assumed to be the realization of state feedback, (2) is shown to be equivalent to the observability constraint. TL is the unobservability subspace and is invariant under A. Finally, (3) is considered to take care of internal stability. Solution to the general problem is then presented. The results are illustrated by regulation in the presence of certain type of disturbances.
  • Thumbnail Image
    Item
    Trajectory planning for robot arm positioning
    (1985) Datta, Subroto; Pearson, J. B.; Figueiredo, Rui J. P. de; Burrus, C. S.; Kohn, Wolf
    This thesis deals with the sub-optimal control of robots. Keeping in mind the current applications of robots in industry we propose a control scheme which minimizes a time-energy cost criterion. The control problem is solved for a piecewise linear model which is sufficiently close in a norm sense to the highly non-linear robot. A solution is obtained by splitting the control problem into two sub-problems. i) A trajectory planning problem which is solved off-line and is termed the NODES problem. The nodes problem primarily obtains the average force history. ii) A feedback controller problem which moves the robot between the successive nodes in real time and is termed the LOCAL problem. The nodes problem is solved using monotonicity ideas arising from dynamic programming and an iterative algorithm is presented for its solution. Numerical issues in the implementation of the algorithm are discussed. The local problem is solved by a modification of a recent paper on minimum time-fuel control.