Junhong Nie

Learning control using fuzzified self-organizing radial basis function network

This note describes an approach to integrating fuzzy reasoning systems with radial basis function (RBF) networks and shows how the integrated network can be employed as a multivariable self-organizing and self-learning fuzzy controller. In particular, by drawing some equivalence between a simplified fuzzy control algorithm (SFCA) and a RBF network, we conclude that the RBF network can be interpreted in the context of fuzzy systems and can be naturally fuzzified into a class of more general networks, referred to as FBFN, with a variety of basis functions (not necessarily globally radial) synthesized from each dimension by fuzzy logical operators. On the other hand, as a result of natural generalization from RBF to SFCA, we claim that the fuzzy system like RBF is capable of universal approximation. Next, the FBFN is used as a multivariable rule-based controller but with an assumption that no rule-base exists, leading to a challenging problem of how to construct such a rule-base directly from the control environment. We propose a simple and systematic approach to performing this task by using a fuzzified competitive self-organizing scheme and incorporating an iterative learning control algorithm into the system. We have applied the approach to a problem of multivariable blood pressure control with a FBFN-based controller having six inputs and two outputs, representing a complicated control structure. >

Constructing fuzzy model by self-organizing counterpropagation network

This paper describes a general and systematic approach to constructing a multivariable fuzzy model from numerical data through a self-organizing counterpropagation network (SOCPN). Two self-organizing algorithms USOCPN and SSOCPN, being unsupervised and supervised respectively, are introduced. SOCPN can be employed in two ways. In the first place, it can be used as a knowledge extractor by which a set of rules are generated from the available numerical data set. The generated rule-base is then utilized by a fuzzy reasoning model. The second use of the SOCPN is as an online adaptive fuzzy model in which the rule-base in terms of connection weights is updated successively in response to the incoming measured data. The comparative results on three well studied examples suggest that the method has merits of simple structure, fast learning speed, and good modeling accuracy. >

Modeling pH neutralization processes using fuzzy-neural approaches

Abstract This paper is concerned with the modeling and identification of pH-processes via fuzzy-neural approaches. A simplified fuzzy model acting as an approximate reasoner is used to deduce the model output on the basis of the identified rule-base which is derived by using one of the following three network-based self-organizing algorithms: unsupervised self-organizing counter-propagation network (USOCPN), supervised self-organizing counter-propagation network (SSOCPN), and self-growing adaptive vector quantization (SGAVQ). Three typical pH processes were treated including a strong acid-strong base system, a weak acid-strong base system, and a two-output system with buffering taking part in reaction. Extensive simulations including on-line modeling have shown that these nonlinear pH-processes can be modeled reasonably well by the present schemes which are simple but efficient.

FCMAC: a fuzzified cerebellar model articulation controller with self-organizing capacity

Abstract The Albuss Cerebellar Model Articulation Controller (CMAC) network has been used in many practical areas with considerable success. This paper presents a fuzzified CMAC network (FCMAC) acting as a multivariable adaptive controller with the feature of self-organizing association cells and the further ability of self-learning the required teacher signals in real-time. In particular, the original CMAC has been reformulated within a framework of a simplified fuzzy control algorithm (SFCA) and the associated self-learning algorithms have been developed as a result of incorporating the schemes of competitive learning and iterative learning control into the system. By using a similarity-measure-based, instead of coding-algorithm-based, content-addressable scheme, FCMAC is capable of dealing with arbitrary-dimensional continuous input space in a simple manner without involving complicated discretizing, quantizing, coding, and hashing procedures used in the original CMAC. The learning control system described here can be thought of as either a completely unsupervised fuzzy-neural control strategy without relying on the process model or equivalently an automatic real-time knowledge acquisition scheme for the implementation of fuzzy controllers. The proposed approach has been applied to a multivariable blood pressure control problem which is characterized by strong interaction between variables and large time delays.

A fuzzy controller with decoupling for multivariable nonlinear servo-mechanisms, with application to real-time control of a passive line-of-sight stabilization system

A fuzzy controller with decoupling suitable for the control of multivariable nonlinear servomechanisms is proposed. By using the paradigm of decentralized control, a control structure is developed which is composed of two control loops, each associated with a single-variable fuzzy controller and a decoupling unit. Real-time experimental results in applying the proposed fuzzy controller with decoupling to a prototype passive line-of-sight stabilization system are presented to demonstrate its effectiveness in a typical application.

A self-organizing fuzzified basis function network control system applicable to nonlinear servomechanisms

The application and extension of suitable techniques for integrating neural network and fuzzy system methodologies to enhance a basic fixed neural network-based nonlinear control strategy with the property of self-organization are investigated. It is shown that by establishing a suitable correspondence between Radial Basis Function networks and fuzzy systems, it is possible to develop a self-organizing controller, utilizing a class of Fuzzified Basis Function Networks (FBFN), that autonomously organizes its network structure to the required size and parameters. The effectiveness of the proposed self-organizing FBFN control system is demonstrated in real-time implementation experiments for position control in a servomechanism with asymmetrical loading and changes in the load.

A fuzzy-neural approach to time series prediction

This paper presents a fuzzy-neural approach to prediction of nonlinear time series. The underlying mechanism governing the time series, expressed as a set of IF-THEN rules, is discovered by a modified self-organizing counter propagation network. The task of predicting the future is carried out by a fuzzy predictor on the basis of the extracted rules. We have applied the approach to three well studied time series. Comparative studies with the other approaches on the sunspot, flour prices, and Mackey-Glass chaotic time series suggest that our approach can offer comparable or even better performances. One of the salient features of the approach is that only single leaning epoch is needed, thereby providing a useful paradigm for some situations where the fast learning is critical.<<ETX>>

A rule-based channel equalizer with learning capability

The problem of channel equalization is concerned with reconstructing binary signal being transmitted through a dispersive communication channel and then corrupted by additive noise. With the aid of fuzzy concepts and neural-like learning, this paper presents a rule-based approach to this problem. A self-organizing algorithm consisting of learning, pruning, and refining processes is developed aiming at building the rule-base from labeled observations. The rule-based equalizer makes the decision on the basis of measuring the similarity between the current observation and the obtained rule prototypes. The simulation studies on linear and nonlinear channels were used to demonstrate the performance of the proposed approach.

Developments in learning control enhancements for nonlinear servomechanisms

Two learning control enhancement schemes applicable to nonlinear dynamical systems of the type commonly encountered in many practical servomechanisms are developed. The key difference between the two schemes lies in the fact that the first scheme utilizes previous cycle errors in the learning strategy while the second utilizes present cycle errors. The uniqueness of these learning controllers is that, in contrast to existing learning methods, they make full use of available a priori information on nominal models of the system and, as a result, overcome the problem of large swings in the control effort during the learning process, which is a major problem with conventional learning controllers. Properties of the schemes are discussed, and it is shown that the strategies developed assure iterative improvement in repetitive operation and asymptotic tracking of the reference signal. The performance and effectiveness of the proposed schemes are then investigated and compared through simulation.

Fuzzy modeling of nonlinear pH processes through neural approach

This paper is concerned with the modeling and identification of nonlinear pH-processes via fuzzy-neural approaches. A simplified fuzzy model acting as an approximate reasoner is used to deduce the model output on the basis of the identified rule-base which is derived by using network-based self-organizing algorithms. Two typical pH processes were treated including a weak acid-strong base system and a two-output system with buffering taking part in reaction. Simulation results have shown that these nonlinear pH-processes can be modeled reasonably well by the present schemes which are simple but efficient.<<ETX>>