Showing papers on "Robustness (computer science) published in 1984"

Analytical redundancy and the design of robust failure detection systems

Abstract: The failure detection and identification (FDI) process is viewed as consisting of two stages: residual generation and decision making. It is argued that a robust FDI system can be achieved by designing a robust residual generation process. Analytical redundancy, the basis for residual generation, is characterized in terms of a parity space. Using the concept of parity relations, residuals can be generated in a number of ways and the design of a robust residual generation process can be formulated as a minimax optimization problem. An example is included to illustrate this design methodology.

Sliding controller design for non-linear systems

Abstract: New results are presented on the sliding control methodology introduced by Slotine and Sastry (1983) to achieve accurate tracking for a class of non-linear time-varying multivariate systems in the presence of disturbances and parameter variations. An explicit trade-off is obtained between tracking precision and robustness to modelling uncertainty : tracking accuracy is sot according to the extent, of parametric uncertainty and the frequency range of unmodelled dynamics. The trade-off is further refined to account for time-dependence of model uncertainty.

An 0(n 2.5 ) Fault Identification Algorithm for Diagnosable Systems

Abstract: Consider a system composed of n independent processors, each of which tests a subset of the others. It is assumed that at most tp of these processors are permanently faulty and that the outcome of a test is reliable if and only if the processor which performed the test is fault free. Such a system is said to be tp-diagnosable if, given any complete collection of test results, the set of faulty processors can be uniquely identified. In this paper, it is shown that tp-diagnosable systems, due to their robust interconnection structure, possess heretofore unknown graph theoretic properties relative to vertex cover sets and maximum matchings. An 0(n2.5) algorithm is given which exploits these properties to identify the set of faulty processors in a tp-diagnosable system. The algorithm is shown to be correct, complete, not based on any conjecture, and superior to any other known fault identification algorithm for the general class of tp-diagnosable systems.

Quadratically Convergent Optimal Power Flow

Abstract: A newly developed sparse implementation of an optimization method using exact second derivatives is applied to the optimal power flow problem. Four utility systems are studied using a variety of objective functions, including fuel costs, active and reactive losses, and new shunt capacitors. Systems solved range from 350 buses to 2000 buses. Comparisons are made with an older algorithm which uses an Augmented Lagrangian to demonstrate the advantages of run time and robustness of the new method. The algorithm and accompanying software represent a technological breakthrough, since they are suitable for solving systems on the order of 2000 buses and demonstrate solution speeds of 5 minutes on large mainframe computers. The method is particularly well suited to infeasible, or even divergent starting points. An option to add shunt capacitors in the event of hopeless infeasibility guarantees an optimal solution for many difficult to solve systems. An automatic scaling feature is added to correct numerical ill-conditioning resulting from series compensation or poor R/X ratios.

Prototyping versus specifying: A multiproject experiment.

Abstract: In this experiment, seven software teams developed versions of the same small-size (2000-4000 source instruction) application software product. Four teams used the Specifying approach. Three teams used the Prototyping approach. The main results of the experiment were the following. 1) Prototyping yielded products with roughly equivalent performance, but with about 40 percent less code and 45 percent less effort. 2) The prototyped products rated somewhat lower on functionality and robustness, but higher on ease of use and ease of learning. 3) Specifying produced more coherent designs and software that was easier to integrate. The paper presents the experimental data supporting these and a number of additional conclusions.

On minimax robustness: A general approach and applications

Abstract: The minimax approach to the design of systems that are robust with respect to modeling uncertainties is studied using a game theoretic formulation in which the performance functional and the sets of modeling uncertainties and admissible design policies are arbitrary. The existence and characterization of minimax robust solutions that form saddle points are discussed through various methods that take into account several common features of the games encountered in applications. In particular, it is shown that if the performance functional and the uncertainty set are convex then a certain type of regularity condition on the functional is sufficient to ensure that the optimal strategy for a least favorable element of the uncertainty set is minimax robust. The efficacy of the methods proposed for a general game is tested in the problems of matched filtering, Wiener filtering, quadratic detection, and output energy filtering, in which uncertainties in their respective signal and noise models are assumed to exist. These problems are analyzed in a common Hilbert space framework and they serve to point out the advantages and limitations of the proposed techniques.

Detection of abrupt changes in dynamic systems

Abstract: In this paper we present some of the basic ideas associated with the detection of abrupt changes in dynamic systems. Our presentation focuses on two classes of methods — multiple filter-based techniques and residual-based methods — and in far more detail on the multiple model and generalized likelihood ratio methods. Issues such as the effect of unknown onset time on algorithm complexity and structure and robustness to model uncertainty are discussed.

Infinitesimal Robustness for Autoregressive Processes

Abstract: We define the influence function and construct optimal robust estimators for autoregressive processes. We show that the asymptotic bias caused by small contaminations of the marginals can be written as the integral of a certain function with respect to the contamination. This function is called the influence function. It is unique only up to an equivalence relation, but there is a natural unique version which describes the limiting influence of an additional observation given the previous observations. Moreover, with this version the asymptotic variance at the true model can be expressed in a simple form. Optimal robust estimators minimize this asymptotic variance under a constraint on the influence function. As in the i.i.d case, they are found by truncating a multiple of the influence function of the maximum likelihood estimator.

Reconstruction from projections based on detection and estimation of objects--Parts I and II: Performance analysis and robustness analysis

Abstract: The problem of reconstructing a multidimensional field from noisy, limited projection measurements is approached using an object-based stochastic field model. Objects within a cross section are characterized by a finite-dimensional set of parameters, which are estimated directly from limited, noisy projection measurements using maximum likelihood estimation. In Part I, the computational structure and performance of the ML estimation procedure are investigated for the problem of locating a single object in a deterministic background; simulations are also presented. In Part II, the issue of robustness to modeling errors is addressed.

Uncertainties in Nuclear Probabilistic Risk Analyses

Abstract: There are many uncertainties in a probabilistic risk analysis (PRA). We identify the different types of uncertainties and describe their implications. We then summarize the uncertainty analyses which have performed in current PRAs and characterize results which have been obtained. We draw conclusions regarding interpretations of uncertainties, areas having largest uncertainties, and needs which exist in uncertainty analysis. We finally characterize the robustness of various utilizations of PRA results.

Robustness of the Two-Sample T-Test

Abstract: In the literature, one finds evidence that the two-sample t-test is robust with respect to departures from normality, and departures from homogeneity of variance (at least when sample sizes are equal or nearly equal). This evidence, presented in various articles, is usually based on an approximate approach without error analysis or on a simulation approach that is of limited extent. The present paper takes a closer and more extensive look at the quality of this procedure under departures from the primary assumptions of normality and of equal variances. The results presented are a synthesis of several previous papers by the author and colleagues, with particular emphasis on the use of a broad Monte Carlo approach to the assessment of robustness.

The Variogram and its Estimation

Abstract: Robustness properties of various variogram estimators are discussed. A closer look at the variogram is made and conditions for the traditional non-parametric estimator to be optimal is presented. Frequently occurring deviations from these conditions are discussed. An alternative robust variogram estimator is defined. In an empirical test, this estimator is found to be more robust towards the deviations than other frequently used variogram estimators.

On the robustness of LQ regulators

Abstract: It is shown that in spite of its large gain and phase margins the linear quadratic state feedback regulator may suffer from poor robustness where small changes in the parameters of the system may lead to fast unstable closed-loop modes.

Model-Reference Adaptive Control—Stability, Parameter Convergence, and Robustness

Abstract: We study stability, parameter convergence and robustness aspects of single input-single output model reference adaptive systems. We begin by establishing a framework for studying parametrizable and unparametrizable uncertainty in the plant to be controlled. Using the standard assumptions on the parametrizable part of the plant dynamics we give a corrected proof (of Narendra, Lin and Valavani) of the stability of the nominal adaptive scheme. Next, we give conditions on the exogenous input to the adaptive loop, the reference signal, to guarantee exponential para meter and error convergence. Using our framework for studying unmodelled (unparametrized) dynamics; we show how the model should be chosen, and the update law modified (by a deadzone in the update law) to preserve stability of the adaptive loop in the presence of output disturbances and unmodelled dynamics. Finally, we compare adaptive and non-adaptive con trol and list directions of ongoing research.

Algebraic theory of linear multivariable feedback systems

Abstract: This paper presents an algebraic theory for analysis and design of linear multivariable feedback systems. The theory is developed in an algebraic setting sufficiently general to include, as special cases, continuous and discrete time systems, both lumped and distributed. Designs are implemented by construction of a controller with two vector inputs and one vector output. Use of controllers of this type is shown to generate convenient stability results, and convenient global parametrizations of all I/O maps and all disturbance-to-output maps achievable, for a given plant, by a stabilizing compensator. These parametrizations are then used to show that any such I/O map and any such disturbance-to-output map may be simultaneously realized by choice of an appropriate controller. In the special case of lumped systems, it is shown that the design theory. can be reduced to manipulations involving polynomial matrices only. The resulting design procedure is thus shown to be more efficient computationally. Finally, the problem of asymptotically tracking a class of input signals is considered in the general algebraic setting. It is shown that the classical results on asymptotic tracking can be generalized to this setting. Additionally, sufficient conditions for robustness of asymptotic tracking, and robustness of stability are developed.

Robustness and modeling error characterization

Abstract: The results on robustness theory presented here are extensions of those given in [1]. The basic innovation in these new results is that they utilize minimal additional information about the structure of the modeling error as well as its magnitude to assess the robustness of feedback systems for which robustness tests based on the magnitude of modeling error alone are inconclusive.

The control of robot manipulators with bounded input: Part II: Robustness and disturbance rejection

Abstract: In this paper we continue the investigations begun in [13] and [14] on the control of robot manipulators. Using the theory of uncertain dynamical systems developed in [4], [5], [16], [17] we derive a robust nonlinear control strategy with guaranteed tracking properties which can be quantified given bounds on the extent of model uncertainty, sensor noise, input disturbances, etc. We also extend the class of pointwise optimal control strategies of [13] to the case of systems with uncertainty in order to treat the problem of input constraints within the context of uncertain systems.

An adaptive array processor with robustness and broad-band capabilities

Abstract: A new type of receiving array which adaptively minimizes ouput noise power while simultaneously satisfying certain robustness and/or bandwidth criteria is considered. The resulting array gains are shown to be robust against direction uncertainty in the assumed look direction, against wavefront distortions and against array geometry errors. The robustness property is incorporated directly into the adaption algorithm via constraints. Extensive simulation has established very satisfactory performance of this new algorithm, both as a limited broad-band processor and as a robust narrow-band processor.

Robust Regression Based on Infinitesimal Neighbourhoods

Abstract: We study robust estimation in the general normal regression model with random carriers permitting small departures from the model. The framework is that of Bickel (1981). We obtain solutions of Huber (1982), Krasker-Hampel (1980) and Krasker-Welsch (1982) as special cases as well as some new procedures. Our calculations indicate that the optimality properties of these estimates are more limited than suggested by Krasker and Welsch.

Robust direct adaptive control

Abstract: This paper proposes a new model reference adaptive control algorithm which has good robustness properties in the presence of unmodeled plant dynamics. The new algorithm requires filtering of the plant input and output with low pass first order filters, prior to their use in the adaptive algorithms. In the case where the dominant transfer function of the plant has a relative degree n* = 1 a global convergence result has been proven in the presence of unmodeled dynamics. When n* > 1 the algorithm employs an adaptive law with normalized signals in order to improve robustness with respect to unmodeled dynamics. It is shown that two of the most crucial factors for robustness, the speed of adaptation and the magnitude of the estimated parameters relative to the speed of the parasitics can be adjusted using the normalized adaptive law.

Robustness of model-reference adaptive control systems with unmodelled dynamics

Abstract: The behaviour of model-reference adaptive control (M R A C) systems in the presence of unmodelled dynamics is analysed. The analysis indicates the existence of robustness, which is confirmed by simulation studies. These results also cast doubt on the conclusions of Rohrs et al. (1982) concerning the lack of robustness in M R A C schemes.

Robustness to Crash in a Distributed Database: A Non Shared-memory Multi-Processor Approach

Abstract: Since attention first turned to the problem of database recovery following system crash, computer architectures have undergone considerable evolution. One direction such evolution has taken is toward fault-tolerant, highly available, distributed database systems. One such architecture is characterized by a single system composed of multiple independent processors, each with its own memory. This paper examines the inadequacy of both the traditional definition of system crash and the conventional approaches to crash recovery for this architecture. It describes an approach to recovery from failures which takes advantage of the multiple independent processor memories and avoids system restart in many cases. This paper appeared in the Proceedings of the Tenth International Conference on Very Large Data Bases, Singapore, Aug. 1984, pp. 445453.