Reinforcement Evolutionary Learning for Neuro-Fuzzy Controller Design
Cheng‐Jian Lin · 2008
A novel reinforcement sequential-search-based genetic algorithm (R-SSGA) is proposed. The better chromosomes will be initially generated while the better mutation points will be determined for performing efficient mutation. We formulate a number of time steps before failure occurs as the fitness function. The proposed R-SSGA method makes the design of TSK-Type fuzzy controllers more practical for real-world applications, since it greatly lessens the quality and quantity requirements of the teaching signals. Two typical examples were presented to show the fundamental applications of the proposed R-SSGA method. Simulation results have shown that 1) the R-SSGA method converges quickly; 2) the R-SSGA method requires a small number of population sizes (only 4); 3) the R-SSGA method obtains a smaller average angular deviation than other methods.