12,000

We have over 12,000 students, from over 100 countries, within one of the safest campuses in the UK

93%

93% of Lancaster students go into work or further study within six months of graduating

Home > Research > Publications & Outputs > Evolving Fuzzy Systems.
View graph of relations

« Back

Evolving Fuzzy Systems.

Research output: Contribution in Book/Report/ProceedingsChapter

Published

Publication date06/2009
Host publicationEncyclopedia of Complexity and System Science
EditorsRobert Mayers
Place of publicationBerlin/Heidelberg
PublisherSpringer
Number of pages10370
VolumeArticl
ISBN (Print)978-0-387-75888-6
Original languageEnglish

Abstract

Fuzzy Sets and Fuzzy Logic were introduced by Lotfi Zadeh in 1965 in his seminal paper [71]. During the last decade of the previous century there was an increase of the various applications of fuzzy logic-based systems mainly due to the introduction of fuzzy logic controllers (FLC) by Ebrahim Mamdani in 1975 [54], the introduction of the fuzzily blended linear systems construct called Takagi–Sugeno (TS) fuzzy systems in 1985 [65], and the the oretical proof that FRB systems are universal approxima tors (that is any arbitrary non-linear function in the [0; 1] range can be asymptotically approximated by a FRB system [68]). Historically, the FRB systems where first being designed based entirely or predominantly on human expert knowledge [54,71]. This offers advantages and was a novel technique at that time for incorporating uncertain, subjective information, preferences, experience, intuition, which are difficult or impossible to be described other- wise. However, it poses enormous difficulties for the pro cess of designing and routine use of these systems, especially in real industrial environments and in on-line and real-time modes. TS fuzzy systems made possible the de- velopment of efficient algorithms for their design not only in off-line, but also in on-line mode [14]. This is facili- tated by their dual nature – they combine a fuzzy linguistic premise (antecedent) part with a functional (usually linear) consequent part [65]. With the invention of the concept of EFS [2,5] the problem of the design was completely automated and data-driven. Thismeans, EFS systems self-develop their model, respectively system structure as well as adapt their parameters “from scratch” on the fly using experimental data and efficient recursive learning mechanisms. Human expert knowledge is not compulsory, not limiting, not essential (especially if it is difficult to obtain in real-time). This does not necessarily mean that such knowledge is prohibited or not possible to be used. On the contrary, the concept of EFS makes possible the use of such knowledge in initialization stages, even during the learning process itself, but this is not essential, it is optional. Examples of EFS are intelligent sensors for oil refineries [52,53], autonomous self-localization algorithms used by mobile robots [75,76], smart agents for machine health monitoring and prognosis in the car industry [30], smart systems for automatic classification of images in CD production process [51] etc. This is a new promising area of research and new applications in different branches of industry are emerging. (c) Springer Verlag

Related research outputs