Akira Ohsumi

Subspace identification for continuous-time stochastic systems via distribution-based approach

This paper presents a novel approach for system identification of continuous-time stochastic state space models from random input-output continuous data. The approach is based on the introduction of random distribution theory in describing (higher) time derivatives of stochastic processes, and the input-output algebraic relationship is derived which is treated in the time-domain. The efficacy of the approach is examined by comparing with other approaches employing the filters.

Nonlinear control of swing-up and stabilization of an inverted pendulum

In this paper an effective method of control for swing-up and stabilization for an inverted pendulum system is established without resorting to any approximation of each nonlinear terms appearing in the mathematical models. The key idea is to derive the partially linearized system by the coordinate change and input transformation via the Lie theoretic approach and to apply a kind of equivalent linearizations to the resulting linear system with nonlinear output injection. Based on the linearized system, the control law is established performing both swing-up and stabilization of the pendulum. The effectiveness of the proposed control method is examined by numerical simulations and tested by experiments.

Active control of flexible structures subject to distributed and seismic disturbances

The purpose of this paper is to present a method of active control for suppressing the vibration of a mechanically flexible cantilever beam which is subject to a distributed random disturbance and also a seismic input at the clamped end. First, the mathematical model of the flexible structure is established by a stochastic partial differential equation which describes the Euler-Bernoulli type distributed parameter system with internal viscous damping and subject to the seismic and distributed random inputs. Second, the distributed parameter model, which is considered as an infinite-dimensional system, is reduced to a finite-dimensional one by using the modal expansion, and split into the controlled part and the uncontrolled (residual) one. The principal approach is to regard the observation spillover due to uncontrolled part as a colored observation noise and construct an estimator, and then we construct the optimal control system. Finally, simulation studies are presented by using a real earthquake accelerogram data.

Subspace-based prediction of linear time-varying stochastic systems

In this paper, a new subspace method for predicting time-varying stochastic systems is proposed. Using the concept of angle between past and present subspaces spanned by the extended observability matrices, the future signal subspace is predicted by rotating the present subspace in the geometrical sense, and time-varying system matrices are derived from the resultant signal subspace. Proposed algorithm is improved for fast-varying systems. Furthermore, recursive implementation of both algorithms is developed.

Instantaneous frequency estimation by using Wigner distribution and Viterbi algorithm

Estimation of the instantaneous frequency (IF) in a high noise environment, by using the Wigner distribution (WD) and the Viterbi algorithm, is considered. The proposed algorithm combines nonparametric IF estimation based on the WD maxima with minimization of the IF variations between consecutive points. Algorithm realization is performed recursively using the (modified) Viterbi algorithm. Performances are compared with IF estimation based on the WD maxima.

Paper: A method of parameter identification for linear distributed parameter systems

A method is presented for estimating unknown parameters in distributed parameter systems. The system considered is assumed to be modeled by a stochastic partial differential equation whose form is known to be linear and unknown parameters are contained in exciting terms. Unknown parameters are assumed to be a set of random constants whose a priori probabilities are known. First, the estimation process of unknown parameters is given by the Bayesian approach in the Markovian framework. The dynamics of the state estimation is also given, which is simultaneously required in the parameter identification scheme. Secondly, the computing procedure is presented, circumventing tedious calculations of the covariance function between the system state and unknown parameters. Finally, two numerical examples are shown, emphasizing that the dynamics of the observation mechanisms adopted plays an important role in both the state estimation and parameter identification.

Online detection of pulse sequence in random noise using a wavelet

Two types of wavelet-based algorithms are proposed for an online detection of a train of unknown pulse signals corrupted by random noise. The mechanism of detecting singularities hidden in the noisy observation data is analyzed, and the performance of the proposed signal detectors is evaluated. Simulation studies are provided to confirm the effectiveness of the algorithms.

Maximum likelihood estimation for signal parameters using pseudo-Wigner distribution

In this paper a novel method for identifying unknown parameters associated with the received signals which are corrupted by random noise is established. Our key approach is to construct a likelihood(-ratio) function for the time-frequency random field generated by the pseudo-Wigner distribution. Numerical simulations are presented to verify the efficacy of the proposed method.

SUBSPACE IDENTIFICATION FOR A CLASS OF TIME-VARYING CONTINUOUS-TIME STOCHASTIC SYSTEMS VIA DISTRIBUTION-BASED APPROACH

Abstract A novel method of subspace identification via distribution-based approach is proposed for a class of time-varying, continuous-time stochastic systems. Efficacy of the method is demonstrated by simulations.

On-line identification of physical parameters involved with dynamic beams: experimental study

An online algorithm for identifying the unknown physical parameters involved with a class of flexible dynamic beams is investigated. The coefficient of the Kelvin-Voigt structural damping and the Youngs modulus of the beam material are considered as unknown parameters. Based on the observation data taken at the free end of the cantilevered beam, the Gauss-Newtonian least squares algorithm is proposed by linking up the iteration number with the current sampling time at which the update observation data is taken. The convergence of the algorithm is demonstrated, and the results of the experiment as well as the simulation studies are presented, emphasising that the simultaneous identification of the damping coefficient and the Youngs modulus is quite important.