http://nova.newcastle.edu.au/vital/access/services/Feed ${session.getAttribute("locale")} 5 http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:11745 In this paper we obtain the maximum likelihood estimate of the parameters of discrete-time state-space models by using a dual time-frequency domain approach. We propose an Expectation Maximization formulation that considers a (non-bijective) linear transformation of the available data. Such a transformation may correspond to different options: selection of time-domain data, transformation to the frequency domain, or selection of frequency-domain data obtained from time-domain samples. We also explore the application of these ideas to Errors-In-Variables systems. 2012-10-16T05:24:46.527Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:11577 Physical systems typically evolve continuously whereas modern controllers and signal processing devices invariably operate in discrete time. Hence sampling arises as a cornerstone problem in essentially all aspects of modern systems science. This paper reviews various aspects of sampling of signals and systems. We argue that careful consideration must be given to sampling to obtain meaningful results when interconnecting a physical system to a computer for the purpose of data storage, signal processing, or control. We also take the opportunity to dispel several common misconceptions about sampling and sampled-data systems. 2012-09-21T03:09:07.339Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:11269 Maximum likelihood estimation has a rich history. It has been successfully applied to many problems including dynamical system identification. Different approaches have been proposed in the time and frequency domains. In this paper we discuss the relationship between these approaches and we establish conditions under which the different formulations are equivalent for finite length data. A key point in this context is how initial (and final) conditions are considered and how they are introduced in the likelihood function. 2012-08-14T06:01:44.728Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:8942 Errors-in-Variables systems have been extensively studied in the literature. We study the impact of sampling on a continuous-time errors-in-variables problem. In particular, we study some approximations of a two dimensional (input-output) continuous-time signal spectrum developed from the sampled-data spectrum. Indeed, some of the paper is tutorial in nature. We also explore the possibility of retrieving the underlying continuous-time system from samples of the input and output signals. 2011-09-14T06:00:13.675Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:8943 Different maximum likelihood formulations have been proposed in the literature for dynamic system identification in the time and frequency domains. In this paper we present numerical examples to study and compare these approaches for short and long data sets. In particular, in the time domain, different likelihood functions are obtained depending on whether or not the initial state is considered as a random vector, as a deterministic parameter, or equal to zero. Similar assumptions can be made in the frequency domain regarding an extra term that contains the difference between the initial and final state. 2011-09-14T06:00:12.792Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:6613 In this chapter we have explored the robustness issues that arise in the identification of continuous-time systems from sampled data. A key observation is that the fidelity of the models at high frequencies generally plays an important role in obtaining models suitable for continuous-time system identification. The problems discussed above have been illustrated for both, deterministic and stochastic systems. Special attention was given to the identification of continuous-time autoregressive stochastic models from sampled data. We have argued that traditional approaches to this problem are inherently sensitive to high-frequency modelling errors. We have also argued that these difficulties can be mitigated by using the proposed FDML with restricted bandwidth. 2010-09-10T01:00:10.359Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:891 The goal of this paper is to contribute to the understanding of fundamental performance limits for feedback control systems. In the literature to date on this topic, all available results assume that the designer has an exact model of the plant. Heuristically, however, one would expect that plant uncertainty should play a significant role in determining the best achievable performance. The goal of this paper is to investigate performance limitations for linear feedback control systems in the presence of plant uncertainty. We formulate the problem by utilizing stochastic embedding of the uncertainty. The results allow one to evaluate the best average performance in the presence of uncertainty. They also allow one to judge whether uncertainty or other properties, e.g., nonminimum phase behavior, are dominant limiting factors. 2010-04-27T06:22:18.484Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:661 Models for deterministic continuous-time nonlinear systems typically take the form of ordinary differential equations. To utilize these models in practice invariably requires discretization. In this paper, we show how an approximate sampled-data model can be obtained for deterministic nonlinear systems such that the local truncation error between the output of this model and the true system is of order Delta(r+1), where A is the sampling period and r is the system relative degree. The resulting model includes extra zero dynamics which have no counterpart in the underlying continuous-time system. The ideas presented here generalize well-known results for the linear case. We also explore the implications of these results in nonlinear system identification. 2010-04-27T05:39:42.083Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:668 This paper addresses the problem of optimal control of constrained linear systems when fast sampling rates are utilised. We show that there exists a well-defined limit as the sampling rate increases. An immediate consequence of this result is the existence of a finite sampling period such that the achieved performance is arbitrarily close to the limiting performance. 2010-04-27T05:39:34.679Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:5195 A precursor to any advanced control solution is the step of obtaining an accurate model of the process. Suitable models can be obtained from phenomenological reasoning, analysis of plant data or a combination of both. Here, we will focus on the problem of estimating (or calibrating) models from plant data. A key goal is to achieve robust identification. By robust we mean that small errors in the hypotheses should lead to small errors in the estimated models. We argue that, in some circumstances, it is essential that special precautions, including discarding some part of the data, be taken to ensure that robustness is preserved. We present several practical case studies to illustrate the results. 2010-04-27T04:47:23.108Z ]]> http://nova.newcastle.edu.au/vital/access/manager/Repository/uon:6042 Most real world systems operate in continuous time. However, to store, analyze or transmit data from such systems the signals must first be sampled. Consequently there has been on-going interest in sampled data models for continuous time systems. The emphasis in the literature to-date has been on three main issues namely the impact of folding, sampled zero dynamics and the associated model error quantification. Existing error analyses have almost exclusively focused on unnormalized performance. However, in many applications relative errors are more important. For example, high performance controllers tend to invert the system dynamics and consequently relative errors underpin closed loop performance issues including robustness and stability. This motivates us to examine the relative errors associated with several common sampled data model types. This analysis reveals that the inclusion of appropriate zero dynamics is essential to ensure that the relative error converges to zero as the sampling period is reduced. 2010-04-27T04:31:49.981Z ]]>