T1: Design of Real Time Evolving Fuzzy Systems: Methodology, Algorithms & Applications

Description

This tutorial summarizes the theoretical, methodological and practical aspects of designing real time evolving fuzzy systems from data. One of the important research challenges today is to design intelligent systems with a higher level of flexibility and autonomy that can develop their understanding of the environment and ultimately their intelligence. It is to be noticed that the environments in which such systems are required to successfully operate are very often ill-defined – they are non-stationary, (often unpredictably) changing, partially or completely unknown. To address the problems of modelling, control, prediction, classification and data processing in such environments a system must be able to fully adapt its structure rather than adjust its parameters based on a pre-trained and fixed structure. That is, the system must be able to evolve, to self-develop, to self-organize. The problem of adaptivity of intelligent systems and their use in on-line mode for real-time applications in industry, defence, advanced technology, biology and medicine has attracted research attention recently. This led during the last few years to the formation of the area of evolving intelligent systems. While conventional adaptive techniques are suitable to represent objects with slowly changing parameters, they can hardly handle complex systems with multiple operating modes or abruptly changing characteristics since it takes a long time after every drastic change in the system to relearn model parameters. The evolving systems paradigm is based on the concept of evolving (expanding or shrinking) model structure that is capable of adjusting to the changes in the objects that cannot solely be represented by parameter adaptation. Evolving intelligent systems develop their structure, their functionality and their internal knowledge representation through continuous learning from data and interaction with the environment. They are synergies between conventional systems, neural networks and fuzzy systems as structures for information representation and the real time methods for machine learning. 

The evolving concept is applicable to a wide range of systems – conventional (multiple model representations), probabilistic (evolvable Bayesian classifiers, decision trees), fuzzy, and neural networks. Cognitive and psychological aspects of the application of evolving paradigm to the system theory identify another of research. Embedded soft computing applications are the natural implementation area of evolving systems as a realistic and practical tool for design of real time intelligent systems. The first part of the tutorial discusses the evolving system methodology within the framework of adaptive, autonomous and self-organizing systems. The second part addresses the main algorithmic and software tools for developing evolving systems from data by addressing the problems of real time structure and parameter learning. The third part focuses on several practical applications of evolving systems in: process control, adaptive speech recognition in VoIP mobile communication, intelligent self-calibrating sensors, machine health monitoring and prognostics, car emission control, system on chip, etc. The tutorial targets control, computer, signal processing, and AI engineers, researchers, practitioners, and graduate students.

About the presenters

Dr. Dimitar P. Filev is a Senior Technical Leader, Intelligent Control and Information Systems with Ford Research and Advanced Engineering specializing in industrial intelligent systems and technologies for control, diagnostics and decision making. He is conducting research in systems theory and applications, modeling of complex systems, intelligent modeling and control and he has published three books, over 160 papers, holds fourteen granted US patents. Dr. Filev is a recipient of the '95 Award for Excellence of MCB University Press and was awarded 4 times with the Henry Ford Technology Award for development and implementation of advanced intelligent control technologies. He is Associate Editor of International Journal of General Systems and of International Journal of Approximate Reasoning. He is a member of the Board of Governors of the IEEE Systems, Man and Cybernetics Society, member of the Fuzzy Systems Technical Committee of the Computational Intelligence Society, IEEE, and the President of the North American Fuzzy Information Processing Society (NAFIPS).

Dr Plamen Angelov, is a member of the Technical Committee on Fuzzy Systems, Chair of a Task Force to the Computational Intelligence Society, and a Senior Member of IEEE. He has more than 100 publications, including a monograph (Springer 2002) and an active research portfolio in the area of computational intelligence, systems and control. He was the Principle Investigator (PI) of a industry-funded projects for development of neural-networks-based approach for electrical load forecasting (1994, NEC), a UK Research Council funded project for development of a methodology for fuzzy rule-based models of system components, a co-Investigator of a research project sponsored by ASHARE-USA. Currently, he is the PI at Lancaster University of ASTRAEA project, a ?27M UK government funded programme, leading the research on Adaptive Routeing and Collision Avoidance for UAVs, the PI of BAE Systems funded project on Collision Avoidance, and several other smaller projects. He is Associate Editor of the Intern. Journal of Knowledge-Based and Intell. Eng. Systems. More information of him can be found on: http://www.lancs.ac.uk/staff/angelov