Title :
Combination of adaptive-network-based fuzzy inference system and incremental best estimate directed search
Author :
Song, Feijun ; Smith, Samuel M.
Author_Institution :
Digital Recorders Inc., Research Triangle Park, NC, USA
fDate :
6/23/1905 12:00:00 AM
Abstract :
Fast incremental best estimate directed Search (fast IBEDS) is a fast and effective search method for Takagi-Sugeno (TS) type fuzzy logic controller (FLC) optimization. Fast IBEDS only optimizes an FLC´s parameters in the rule output functions. The rule antecedents remain fixed during optimization. The paper proposes to combine the adaptive-network-based fuzzy inference system (ANFIS) and fast IBEDS to gain full range optimization of an FLC. The role of ANFIS is to optimize the rule antecedents so that an FLC can be optimized even further. Simulation results on a 4D inverted pendulum show that by combining ANFIS and fast IBEDS together, an FLC can be further optimized to have better global performance
Keywords :
fuzzy control; fuzzy logic; inference mechanisms; iterative methods; optimal control; optimisation; pendulums; position control; search problems; 4D inverted pendulum; ANFIS; Takagi-Sugeno type fuzzy logic controller optimization; adaptive-network-based fuzzy inference system; best kept policy; cell state space performance measures; full range optimization; global performance measures; incremental best estimate directed search; initial best estimate rule; initial best estimate training set; iterative procedure; least mean square algorithm; local performance measures; random values; rule antecedents; Automatic control; Control systems; Convergence; Fuzzy logic; Fuzzy systems; Iterative algorithms; Least squares approximation; Optimization methods; Search methods; State-space methods;
Conference_Titel :
Fuzzy Systems, 2001. The 10th IEEE International Conference on
Conference_Location :
Melbourne, Vic.
Print_ISBN :
0-7803-7293-X
DOI :
10.1109/FUZZ.2001.1007331
