Mahardhika Pratama, Anavatti Sreenatha, Matthew Garrat, Edwin Lughofer,
"Online Identification of Complex Multi-Input-Multi-Output System Based on Generic Evolving Neuro-Fuzzy Inference System"
: Proceedings of the IEEE SSCI 2013 Conference, Serie IEEE SSCI 2013 Conference, IEEE, Seite(n) 106-113, 4-2013
Original Titel:
Online Identification of Complex Multi-Input-Multi-Output System Based on Generic Evolving Neuro-Fuzzy Inference System
Sprache des Titels:
Englisch
Original Buchtitel:
Proceedings of the IEEE SSCI 2013 Conference
Original Kurzfassung:
Unmanned Aerial Vehicles (UAV) has been
deployed for miscellaneous defence operations and commercial
civilian applications. Nowadays, identification of the UAV
dynamic elicits overwhelming interests within the community.
This is mainly inflicted by the pivotal characteristic of the UAV
suffering from the Multi Input Multi Output (MIMO), nonlinear,
non-stationary and erratic natures which are attractive
to be explored. In essence, the identification of the UAV is
time-critical in which a computationally prohibitive algorithm
is unappealing to be taken place. Unfortunately, the
omnipresent approach in identifying the dynamic of the UAV
still relies on offline or batched learning procedure
necessitating a complete training set to be available a priori.
This rationale unwraps a new unchartered territory of
Evolving Neuro-Fuzzy System (ENFS) which is well-known
efficient learning machine capable of coping with any
variations of the system dynamic. Nonetheless, the ENFS is
capable of commencing its learning process from an empty rule
base or initial trained fuzzy model which is alike with the
autonomous mental development in human?s brain. For
brevity, the online identification strategy of the UAV based on
a novel ENFS namely GENEFIS which is a short form of
Generic Evolving Neuro-Fuzzy Inference System is addressed
by this paper. In summary, our proposed algorithm is not only
usable to online identification of the UAV but also can
outperform the state of the art algorithms in terms of
predictive quality and compactness of the rule base.
Forschungs-