Input-output Information Research Articles

A new nonlinear robust disturbance observer

This paper presents a new nonlinear robust disturbance observer by using only the input–output information for minimum phase dynamical systems with arbitrarily relative degrees. The model uncertainties, the nonlinear parts of the system are merged into the disturbance term, and are regarded as a part of the disturbance. The disturbance is assumed bounded. However, the upper and lower bounds are assumed unknown. The new disturbance observer is inspired by the VSS control theory, whereas the upper and lower bounds of the disturbance are adaptively updated. The estimation error of the disturbance can be made as small as necessary by choosing the design parameters. The attraction of the proposed method lies in its robustness to disturbance and easiness to be implemented. Example and simulation results show the effectiveness of the proposed algorithm.

Stochastic Resonance in the Information Capacity of a Nonlinear Dynamic System

We consider a nonlinear bistable dynamic system governed by the quartic potential with two-state quantization at the output — the earliest system to have revealed the phenomenon of periodic stochastic resonance. We devise a scheme in which this system is used to transmit a broadband aperiodic informative signal. With this scheme, we demonstrate that the system can be operated as a memoryless symmetric binary channel, and we develop the characterization of the transmission up to the evaluation of the input–output information capacity of this channel. We show that a regime exists where the information capacity can be increased by means of noise addition, a property we interpret as a form of aperiodic stochastic resonance. In addition, we demonstrate that a positive input–output gain in the efficacy of the signal recovery can be obtained with the stochastic resonator, compared to the recovery that would directly operate on the input signal-plus-noise mixture.

Modeling and control of an eddy current brake

The purpose of this paper is the application of some existing techniques for the modelling and regulation of an eddy current brake process. An approximate theoretical model is derived for the behavior of an eddy current disc brake in the low-speed zone. Input-output information is used to obtain a polynomial state-affine behavior model for such a process. Dynamic and static feedback compensator schemes are proposed for the process speed control in the presence of unknown braking resistant torque. The dragging torque value is estimated by an appropriate nonlinear observer. Experimental results are presented.

Neural classifiers for dynamic system modes

In this study we consider the use of neural networks for system identification. No a priori information is assumed about the system model. Rather it is assumed that the system has several modes of behavior and that a sequence of input-output pairs, while the system is in each mode, is available. Using the input-output information, a neural network is designed with the goal of recognizing each of the modes. The design is tested by using the input-output sequence of an unknown variable system, whose characteristics transition between those of the five prototypes. The identification procedure operates on-line. The ability to identify and track the transitions of the unknown system is explored.

Study on Artificial Neural Networks by Introduction of Magnetic Field. A Unit Hybrid Artificial Neuron Model and Its Application.

In this paper, we propose a unit hybrid artificial neuron (U-Hynuron) model that introduces magnetic field information into artificial neural networks. The U-Hynuron model differs greatly from conventional neural models in two ways : one is that it treats two distinct types of input-output information ; and the other is tthat it is a three-dimensional model using field theory as its mathematical method. A mathematical memory-learning relation has been derived from this model, and is called the learning rule. By this learning rule, reflex learning and union learning are defined. Furthermore, the reflex learning theorem and union learning theorem are proposed and proven. Finally, as an application of pattern recongnition, we solve a recognition problem of Japanes paper money by using U-Hynuron.

NEURAL NETWORKS AS OPTIMAL INFORMATION PROCESSORS

We explore the properties of a feed-forward neural network whose couplings are chosen in such a way as to maximize the input-output mutual information, in the case in which the input-output channel is affected by noise.

Model reduction of unstable systems

A simple and powerful method for unstable model reduction has been developed in which the approach is based on the fact that translation transformations in the s-plane preserve the input-output properties of a system. Using translation transformations in the frequency domain it is possible to change the stability of the system without losing input-output information. Although balancing requires that the model be asymptotically stable, it reduces the model depending only on the information of input to state and state to output. The stability requirement comes from the computation of the controllability and observability gramians which are used for characterizing the contribution of the states to the input-output map. It has been shown in this paper that it is possible to use balancing to reduce the models of unstable systems by transforming them into the stable models, reducing the model order, and then transforming the models back. The method has been demonstrated by case studies.

Augmented model-based double iterative loop techniques for hierarchical control of complex industrial processes

Two novel hierarchical structures are presented which extend the applicability of previous model-based double iterative loop techniques using input-output information feedback to non-convex problems. The techniques incorporate integrated system optimization and parameter estimation, which utilize process measurements to achieve real process optimality in spite of model reality differences. The double iterative loop structures of the proposed algorithms use the real process measurements within the output loops while the inner loops involve model-based computations only. By this means the algorithms use available information from the real process efficiently and a significant reduction in set-point alterations to real sub-processes is achieved. In order to cater for process complexity, the inner loops are organized in the form of two-level hierarchical structures. An applicability of the techniques to non-convex problems is achieved by a suitable convexification of an original process performance index. It ...

Optimal Information Processing and Bayes's Theorem

Abstract In this article statistical inference is viewed as information processing involving input information and output information. After introducing information measures for the input and output information, an information criterion functional is formulated and optimized to obtain an optimal information processing rule (IPR). For the particular information measures and criterion functional adopted, it is shown that Bayes's theorem is the optimal IPR. This optimal IPR is shown to be 100% efficient in the sense that its use leads to the output information being exactly equal to the given input information. Also, the analysis links Bayes's theorem to maximum-entropy considerations.

Macroeconomic structure and computable general equilibrium models

Economy-wide analysis can be undertaken within three different accounting frameworks: (1) the national income and product accounts, which focus on balance among macroaggregates; (2) the input-output accounts, which focus on intermediate flows and on the sectoral composition of production and demand; and (3) the social accounting matrix (SAM), which provides a framework incorporating both national income and product as well as input-output information. All three of these frameworks have been used to develop fixed coefficient, linear multiplier models. Keynesian multipliers at the macro level have multisectoral counterparts in the input-output model and “multi-institutional” counterparts in SAM-based models. These linear multiplier models tend to be completely demand driven and do not incorporate supply constrains or substitution possibilities. Computable general equilibrium (CGE) models have been developed which capture nonlinear substitution possibilities and multisectoral supply-demand interactions, and also incorporate macro variables and mechanisms for achieving balance among aggregates. “Marginal” multipliers can be derived for CGL models using the Jacobian matrix of partial derivatives at a given equilibrium. This paper develops an approach to analyzing Jacobian multipliers from a CGE model that decomposes macro, sectoral, and institutional linkages in a SAM framework. We give an illustration of SAM-multiplier decomposition using a small CGE model of the United States.

Modeling of nonlinear discrete-time systems from input-output data

An algorithm is presented that produces a polynomial state affine model for a discrete-time nonlinear system. The method uses only input-output information and is based on a difference equation approximation to the input-output map. From the difference equation, a behaviour matrix is constructed, to be used in the state affine realization algorithm. The difference equation approach requires the estimation of fewer parameters than the usual Volterra series. A state space model is then obtained directly from the difference equation. As a byproduct, a method for calculating the Volterra series directly from the difference equation is obtained. Practical applications to nonlinear systems are included.

Linear compensator designs based exclusively on input-output information are never robust with respect to unmodeled dynamics

The effects of unmodeled, higher-order dynamics or parasitics on the stability of linear control systems are investigated. A description is first given of a class of perturbations of a given state equation that cannot be distinguished from the original on the basis of input-output measurements alone. Then it is shown that given any plant-compensator pair, perturbations of each system of this type can always be found that destabilize the closed-loop configuration. The effect of destabilizing perturbations on output behavior is explored. >

Discrete-time decentralized adaptive control

A discrete-time decentralized model reference adaptive controller is presented for a large-scale system composed of N interconnected single-input single-output linear subsystems with unknown parameters and unmodeled dynamics. Each local adaptive law utilizes a normalizing signal which is generated by using information of the delayed inputs and outputs of other subsystems. The delay d in exchanging input-output information between subsystems is required to be finite for global stability and robustness. However, this delay may be crucial for performance since a large d leads to a larger bound for the residual tracking error.

Modeling of Nonlinear Discrete Time Systems from Input-Output Data

An algorithm is presented that produces a polynomial state affine model for a discrete time nonlinear system. The method uses only input-output information and is based on a difference equation approximation to the input-output map. From the difference equation, a behavior matrix is constructed, to be used in the state affine realization algorithm. The difference equation approach requires the estimation of fewer parameters than the usual Volterra series. A state space model is then obtained directly from the difference equation. As a by-product, a method for the calculation of the Volterra series from the difference equation is obtained.

The Effect of Reductions in Concentration on Income Distribution

The simulation model of this study indicates that a decline in above n ormal profits associated with concentration will cause a redistribution of incom e from the highest of six income classes to low and middle income classes. The m agnitude of gains and losses range from 0.2 to 0.9 percent of income for reducti ons in four-firm concentration ratios to 50 percent in all manufacturing industr ies. The methodology uses consumer expenditure data and input-output information to estimate the impact on payments made to capital, and uses income tax data on income sources by income class to estimate the impact on income received from c apital ownership. Estimates of the change in profits resulting from a change in concentration are based on a published concentration-profits regression. Copyright 1987 by MIT Press.

Conditional entropy in algebraic statistical physics

The notion of conditional entropy as entropy of conditional state on C ∗-algebra A with respect to its C ∗-subalgebra A 1⊂ A is introduced. It is proved that for a compatible state σ on A (which admits the conditional expectation of Umegaki-Takesaki) the mean conditional entropy in an a priori state σ 1 on A 1 is equal to the difference of the entropy of the state σ on A and the entropy of the state σ 1 on A 1. The conditional entropy enables us to define the input-output information of a quantum communication channel in analogy to the classical Shannon formula.

State Observation from Quantized Input-Output Information

Systems whose control inputs and observation outputs can take only discrete-level values are called quantized-control Quantized-observation systems or simply quantized systems. In this paper controllability and observability properties of quantized systems and their relationship to system structure are investigated for linear, finite-dimensional and discrete-time systems. It is shown that every Quantized control system admits a canonical decomposition into three subsystems corresponding to the states which are discretely reachable, almost reachable and almost reachable in finite time. This canonical control structure explicitly shows to what extent a system can be controlled using Quantized control inputs. With the help of this structure theorem the problems of state estimation from Quantized input-output information are discussed and some conditions for a quantized system to be observable are obtained in terms of the coefficient matrices of the system equations.

Effects of instruction, visual imagery and educational background on process control performance

The effects of instruction, vividness of visual imagery and education were evaluated on operator control performance on a relatively complex process simulator. Thirty-two subjects took part. Analysis of variance of the data indicated a significant effect due to instruction. No significant effects were found for vividness of visual imagery or educational background, nor any significant interactions. A significant learning effect was shown to exist for one of the levels of instruction, namely the group with process information. The group with only input-output information did not show any learning effect over the three experimental trials.

Extended Levinson and Chandrasekhar equations for general discrete-time linear estimation problems

Recursive algorithrms for the solution of linear least-squares estimation problems have been based mainly on state-space models. It has been known, however, that recursive Levinson-Whittle-Wiggins-Robinson (LWR) algorithms exist for stationary time-series, using only input-output information (i.e, covariance matrices). By introducing a way of classifying stochastic processes in terms of an index of nonstationarity we derive extended LWR algorithms for nonstationary processes We show also how adding state-space structure to the covariance matrix allows us to specialize these general results to state-space type estimation algorithms. In particular, the Chandrasekhar equations are shown to be natural descendants of the extended LWR algorithm.

Automatic insertion of indexing instructions in program synthesis

A algorithm is described for creating a computer program from a trace of an example calculation which may omit certain indexing operations. The technique is designed to be used in an interactive programming system that allows a user to work example computations at a display console in scratchpad fashion and that automatically synthesizes a program which can do the computations. A number of programs are given that the algorithm was used to synthesize. Finally, it is shown that this approach yields a technique for synthesizing certain computer programs from input-output information only.