A simplified architecture for triangular quasi type-2 fuzzy logic systems

Abstract

Type-2 fuzzy logic systems (T2 FLSs) are known for modelling and handling high levels of uncertainties associated with most of the real world applications. However, their very high computational complexities have generally prevented their wide spread application. Objectives of this paper are to: 1) present a simplified architecture for implementing a triangular quasi T2 FLS (QT2 FLS) that uses three embedded type-1 FLSs; 2) tune the footprint of uncertainty (FOU) with an objective to optimise the performances of interval T2 and QT2 FLSs; 3) compare the performances of T1, IT2 and triangular QT2 FLS (based on the proposed architecture). For validating the proposed architecture, it has been applied for forecasting of Mackey-Glass Time-Series. T1 FLS is firstly evolved using particle swarm optimisation (PSO) algorithm and is then upgraded to QT2 FLS. The simulation results for different noise levels and different FOUs show that the QT2 FLS based on the proposed realisation approach, can handle higher levels of data uncertainties under noisy conditions compared to IT2 FLS, and hence outperforms its T1 and IT2 counterparts.