Piotr Kozierski

Stability analysis and tracking performance of fractional-order PI controller for a second-order oscillatory system with time-delay

The main aim of this paper is to present the analysis of stability regions and tracking performance for the closed-loop system with second-order plant with time-delay and continuous fractional-order PI controller. It is a trial to extend the results and methodology presented in [11] to non-integer order systems. The tracking performance is based on two quality indices IAE and ISE. The tuning method of fractional PI controller is with respect to Hermite-Biehler and Pontryagin Theorems.

Stability analysis of simple anti-windup compensation in approximate pole-placement control of a second order oscillatory system with time delay

The main aim of this paper is to present analysis of anti-windup compensation impact on tracking performance for a second-order plant with time delay and continuous-time PID controller tuned by pole-placement. The paper uses a first-order non-Pade approximation of the delay in order to streamline the solution of the Diophantine equation. In the case of using Pade approximation, the solution could only be obtained approximately, because of greater number of variables than equations. The paper also studies influence of time delay approximation with first-order model to make approximate pole-placement possible. The performance of the system is compared on the basis of computing differences between integral of absolute and squared tracking errors for the system with and without compensation, as well as observing the excess of windup phenomenon taking place. The control system quality is described on the basis of two quality indices in the case of stable oscillatory second-order plant and square-wave reference signal.

Tracking control of an inertial wheel pendulum by LQR regulation

This article describes universal tracking control method. It is based on model linearization in every point of trajectory. Tested device is an inertial wheel pendulum (IWP). This is an underactuated nonlinear object - two degrees of freedom (angle from vertical and angle of rotation of electric motor) and one actuator (current). The linear quadratic regulator (LQR) and the model of the object are analytically considered. Many experiments are presented and confirm proper and stable operation of the system and some constraints.

Dispersed filters for power system state estimation

The article proposes an approach to power system state estimation allowing the division of the network into smaller parts and performing calculations for each part at the same time. The latter can be implemented in parallel, but the main aim has been to propose a method for dispersed calculations, i.e. calculations that may be performed on computing units located at various points of the whole power system. In the paper, there are 3 algorithms for which dispersed versions have been proposed: Extended Kalman Filter, Particle Filter and Extended Kalman Particle Filter. As a result of the simulations, it has been verified that the Dispersed Particle Filter works better than simple Particle Filter. In two other cases, distributed algorithms work worse, but for the Extended Kalman Filter degradation in the estimation quality is not significant.

Particle Filter in State Vector Estimation Problem for Power System

Particle Filter is a tool, which has been used more frequently over the years. Calculations with using Particle Filter methods are very versatile (in comparison to the Kalman Filter), which can be used in high complex and nonlinear problems. Example of such a problem is the power system, where Particle Filter is used to state estimation of network parameters based on measurements. Paper presents theoretical basis regarding Par- ticle Filter and power system state estimation. Results of experi- ment have shown that Particle Filter usually gives better outcome comparing to the Weighted Least Squares method. In extension Multi Probability Density Function Particle Filter is proposed, which improves obtained results so that they are always better than Weighted Least Squares method.

Kaldi Toolkit in Polish Whispery Speech Recognition

In this paper, the automatic speech recognition task has been presented. Used toolkits, libraries and prepared speech corpus have been described. The obtained results suggest, that using different acoustic models for normal speech and whispered speech can reduce word error rate. The optimal training steps has been also selected. Thanks to the additional simulations it has been found that used corpus (over 9 hours of normal speech and the same of the whispery speech) is definitely too small and must be enlarged in the future. Streszczenie. W artykule przedstawiono automatyczne rozpoznawanie mowy. Wykorzystane narzędzia, biblioteki i korpus opisano w artykule. Uzyskane wyniki wskazują, że wykorzystując różne modele akustyczne dla mowy zwykłej i szeptanej uzyskuje się polepszenie skuteczności rozpoznawania mowy. W wyniku wykonanych badań wskazano również optymalną kolejność kroków treningu. Dzięki dodatkowym obliczeniom stwierdzono, że użyty korpus (ponad 9 godzin zwykłej mowy i drugie tyle szeptu) jest zdecydowanie za mały do dobrego wytrenowania systemu rozpoznawania mowy i w przyszłości musi zostać powiększony. (Narzędzia Kaldi w rozpoznawaniu polskiej mowy szeptanej).

Particle filter in multidimensional systems

The article presents studies on the estimation quality of a particle filter applied to small multidimensional objects. For the purposes of the article, a new type of network has been proposed, in which each node is associated with one state variable. Based on performed simulations it has been found that particle filter implemented for small systems (1- or 2-dimensional) is a good choice. However, for larger objects Kalman filter may return better results (it depends on the chosen particles number). This is due to the exponential dependence of needed particles number to the object dimension. It has been also observed that the particle filter, in comparison to the Extended Kalman filter, better estimates the state variables which are well metered, and simultaneously worse estimates the state variables which are worse metered. Possible approaches for objects with a greater number of state variables also have been adduced, including the dispersed particle filter.

Fractional-Order PI Controller with Anti-windup Compensation for First Order Delay System

The main aim of this paper is to present the analysis of stability regions for the closed-loop system with first-order linear model of the Inteco Servo Drive with additional time delay and continuous fractional-order PI controller and anti-windup compensators. It is a extension to Authors previous works concerning non-integer order controllers, in which stability issues are considered by means of BIBO simulation-based tests, with several anti-windup compensation schemes considered. The tuning method of fractional-order PI controller is with respect to Hermite-Biehler and Pontryagin theorems.

Particle Filtering in Servo Drive Velocity Control with Fractional-Order PI Controller

The main aim of this paper is to compare the closed-loop system based on combination of Fractional-order PI and Particle Filter with combination of Fractional-order PI and Kalman Filter, based on quality IAE index. Authors verified the results of a real-world experimental setup which was the Modular Servo System. The article is a combination of previous work and implementation of fractional-order PI controller with Particle Filter or Kalman Filter to a real-world system.

The impact of vocabulary size and language model order on the polish whispery speech recognition

The article presents studies on the automatic whispery speech recognition. In the performed research a new corpus with whispery speech has been used. The aim of studies presented in this paper was to check, how the vocabulary size and the language model order influence on the speech recognition quality. It has been concluded that even using recordings with 5,000 different words only it is possible to prepare large vocabulary continuous speech recognition (LVCSR) model. It has been also found that the third order of language model is the best choice. The difference between normal and whispery speech is negligible and is manifested only in higher word error rate index (about 1.5 times higher for whispery speech).