Juan C. Agüero

Identifiability of errors in variables dynamic systems

There has been substantial research carried out on the errors in variables (EIV) identifiability problem for dynamic systems. These results are spread across a significant volume of literature. Here, we present a single theorem which compactly summarizes many of the known results. The theorem also covers several cases which we believe to be novel. We analyze single input single output systems using second order properties. We also extend the results to a class of multivariable systems.

On identification of FIR systems having quantized output data

In this paper, we present a novel algorithm for estimating the parameters of a linear system when the observed output signal is quantized. This question has relevance to many areas including sensor networks and telecommunications. The algorithms described here have closed form solutions for the SISO case. However, for the MIMO case, a set of pre-computed scenarios is used to reduce the computational complexity of EM type algorithms that are typically deployed for this kind of problem. Comparisons are made with other algorithms that have been previously described in the literature as well as with the implementation of algorithms based on the Quasi-Newton method.

Choosing Between Open- and Closed-Loop Experiments in Linear System Identification

This correspondence shows that open-loop experiments are optimal for a broad class of systems when the system input is constrained. In addition, we show that, for a general class of systems, when the output power is constrained, closed-loop experiments are optimal. Both results use a strong notion of optimality and use expressions for estimation accuracy which are nonasymptotic in model order but asymptotic in data length.

Approximate EM Algorithms for Parameter and State Estimation in Nonlinear Stochastic Models

Due to the availability of rapidly improving computer speeds, industry is increasingly using nonlinear process models in calculations that appear further down the control hierarchy. Indeed, nonlinear models are now frequently used for real-time control calculations. This trend means that there is growing interest in the availability of high speed state and parameter estimation algorithms for nonlinear models. One family of algorithms that can be used for this purpose is based on the, so called, Expectation Maximization Scheme. Unfortunately, in its basic form, this algorithm requires large computational resources. In this paper we review the EM algorithm and propose several approximate schemes aimed at retaining the essential flavour of this class of algorithm whilst ensuring that the computations are tractable. We will also compare the EM algorithm with several simpler schemes via a number of examples and comment on the trade-offs that occur.

Sampling and Sampled-Data Models: The Interface Between the Continuous World and Digital Algorithms

Modern signal processing and control algorithms are invariably implemented digitally, yet most real-world systems evolve in continuous time. Hence, the interaction between sampling and the behavior of continuous-time systems is an important ingredient in all real-world signals and systems problems.

Sampling and sampled-data models

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.

On the equivalence of time and frequency domain maximum likelihood estimation

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.

ROBUST IDENTIFICATION OF PROCESS MODELS FROM PLANT DATA

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.

Dual time-frequency domain system identification

In this paper we obtain the maximum likelihood estimate of the parameters of discrete-time linear models by using a dual time-frequency domain approach. We propose a formulation that considers a (reduced-rank) 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 use the proposed approach to identify multivariate systems represented in state-space form by using the Expectation-Maximisation algorithm. We illustrate the benefits of the approach via numerical examples.

Passivity-based control for multi-vehicle systems subject to string constraints

In this paper, we show how heterogeneous bidirectional vehicle strings can be modelled as port-Hamiltonian systems. Analysis of stability and string stability within this framework is straightforward and leads to a better understanding of the underlying problem. Nonlinear local control and additional integral action is introduced to design a suitable control law guaranteeing l 2 string stability of the system with respect to bounded disturbances.