ISSN:0020-7179    

DOI:10.1080/00207179208934314

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

In this article, we are concerned with neural-nets which canlearnto control systems in accordance with a guiding intent, and can alsolearnhow to formulate that control strategy or intent. The overall task of systems control is viewed as being carried out by four components, these being the predictive monitoring net, the control action generator net, the objective function net and the optimization net. This approach and perspective are described and illustrated in this article. In our examples, we show that systems identification can indeed be achieved in the presence of noise and that optimal control can be formulated in a learning mode, by neural nets.

ISSN:0020-7179    

DOI:10.1080/00207179208934315

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

The paper discusses the importance of interpolating memories for different lay-outs of learning control loops, design considerations for such memories and the convergence of the learning process for one such memory. Finally, applications of learning control loops and future research in this field are considered.

ISSN:0020-7179    

DOI:10.1080/00207179208934316

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

Many real-world systems exhibit complex nonlinear characteristics and cannot be treated satisfactorily using linear systems theory. A neural network which has the ability to learn sophisticated nonlinear relationships provides an ideal means of modelling complicated nonlinear systems. This paper addresses the issues related to the identification of nonlinear discrete-time dynamic systems using neural networks. Three network architectures, namely the multi-layer perceptron, the radial basis function network and the functional-link network, are presented and several learning or identification algorithms are derived. Advantages and disadvantages of these structures are discussed and illustrated using simulated and real data. Particular attention is given to the connections between existing techniques for nonlinear systems identification and some aspects of neural network methodology, and this demonstrates that certain techniques employed in the neural network context have long been developed by the control engineering community.

ISSN:0020-7179    

DOI:10.1080/00207179208934317

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

Taking account of the fuzzy nature of human decision making processes and real-time properties, this paper establishes a unified approximate reasoning model based on possibility theory rather than on relational matrix computation. Both fuzzy and random uncertainties can be coped with in the model. In the case of sensor-based situations, a simpler reasoning scheme is derived by introducing the concepts of matching measures. The proposed models may provide another possibility for on-line reasoning in real-time expert system applications. In Part 1, the theoretical foundations are provided, while in Part 2 their application to multivariate blood pressure control is described.

ISSN:0020-7179    

DOI:10.1080/00207179208934318

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

The second part of this paper describes a decentralized fuzzy controller structure for dealing with the multivariable control of human blood pressure. It consists of rule-based fuzzy controllers and a simple compensator unit. The reasoning algorithms used by the fuzzy controllers are based on the unified approximate reasoning model derived by the authors in Part 1. The problem involves two cases: simultaneous control of the arterial pressure and systemic venous pressure, and simultaneous regulation of arterial pressure and cardiac output. Eight reasoning algorithms are chosen for comparisons which are based upon the control performance and the performance robustness. A number of simulation results show that the blood pressure can be controlled successfully by the proposed controller despite the presence of strong interactive effects between the variables. In addition, some useful conclusions about reasoning methods are drawn from the comparative studies.

ISSN:0020-7179    

DOI:10.1080/00207179208934319

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

Instead of seeking a structure mapping from a fuzzy reasoning system to a neural network, this paper is intended to find a functional mapping from a fuzzy logic-based algorithm to the network-based approach. By viewing the given rule-base as defining a global linguistic association constrained by fuzzy sets, approximate reasoning is implemented here by a Backpropagation Neural Network (BNN) with the aid of fuzzy set theory. By paying particular attention to the generalization capability of the BNN, the underlying principles have been examined in detail using two examples: a small demonstration at the linguistic level, and a more realistic problem of multivariable fuzzy control of blood pressure. The simulation results not only indicate the feasibility of the BNN-based approach, but also reveal some deeper similarities which exist in the two methods, which may have some important implications for future studies into fuzzy control. In addition, this work may be considered as another application example of the BNN in the case of continuous outputs and on a relatively larger scale (in the second example the BNN has 26 inputs and 13 outputs, with a total of 2013 weights and thresholds).

ISSN:0020-7179    

DOI:10.1080/00207179208934320

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

To achieve self-organizing control based on fuzzy rules, we propose a fuzzy associative memory system called FAMOUS (Fuzzy Associative Memory Organizing Units System). FAMOUS simulates the knowledge representation and inference process by using fuzzy notation and by association in neural networks. FAMOUS's learning algorithm uses training steps to generate operation skills by modifying the expert knowledge that is initially built-in. A set of fuzzy if-then rules is used for controlling variable parameter processes. The control knowledge is represented as pairs consisting of a ‘condition’ in the if-part and an ‘operation (controller)’ in the then-part. The controllers are designed for optimization and stabilization in specific conditions. The fuzzy controller described in FAMOUS recalls well-trained controllers associated with the input condition and makes the final control output by synthesizing the intermediate outputs of their controllers. FAMOUS can highly refine knowledge by using neural network learning algorithms. In the if-part of the knowledge pairs, the membership function is automatically generated from input data by an unsupervised learning algorithm. In the then-part, each controller is individually trained to perform optimally under a specific condition and to satisfy the constraints of stabilization. To check that the whole controller stabilizes the parameter variance process, we also discuss how to obtain the class of stabilizers. Finally, we apply the proposed method to the control of a small helicopter (which is a variable parameter process) and show its usefulness in designing the controller.

ISSN:0020-7179    

DOI:10.1080/00207179208934321

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

Fuzzy controllers may be either static systems, which have a fixed rule base, or adaptive systems, which have the ability to alter their rules. A discussion of adaptive fuzzy controllers and a comparison with corresponding algebraic techniques concludes that all previous adaptive fuzzy controllers have been of the direct adaptive type. Such controllers use observations of closed loop performance to manipulate the controller rule base directly without any intermediate process model being produced. In this paper, an indirect adaptive fuzzy controller is proposed where an intermediate process model, identified from observed data, is used to perform on-line controller design. The resulting separation of the adaptation system from controller design enables learning convergence to be investigated. Examples are given of both fuzzy model identification and controller design for linear and nonlinear processes.

ISSN:0020-7179    

DOI:10.1080/00207179208934322

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

This paper formulates a new approach to the classical learning/adaptive control. problem. Our approach is based on two key observations: (1( the inherent conflict between control and identification as they compete for the only available resource, namely the input to the plant; (2) when designing and optimizing the performance of a control system the current task, as well as the repertoire of other typical future tasks which the system may encounter during its life time, should be considered. Our approach is formulated for a general nonlinear time-varying plant; thus,unlikeexisting adaptive control theory, the theory for a linear time-invariant system evolves as a special case of the general case. The design for the full lifetime of the system creates a methodology that specifies what current actions should be taken in addition to the tracking of the current reference trajectory, at the expense of some performance degradation in the current task, so as to improve the performance of future tasks: this is the learning trade off. The conflicting objectives, namely, tracking versus learning and current task versus future tasks, are most naturally posed and partially solved in the domain of ‘multiple objective optimization theory’. We demonstrate for linear time-invariant plants with quadratic cost, that Pareto optimal learning adaptive controllers may be obtained by simple ‘out of loop’ mixing, where a scalar controls the tracking versus learning trade off in a reliable way.

ISSN:0020-7179    

DOI:10.1080/00207179208934323

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor