Development of feedback error learning strategies for training neurofuzzy controllers on-line

Neurofuzzy model-based controllers have been successfully applied in practice. This paper reviews the feedback error learning strategies used for training neurofuzzy controllers online. The objective is to identify the weaknesses of existing algorithms. A variation of the feedback error learning strategy, capable of overcoming these limitations, is then proposed. Simulation results are presented to show that the proposed feedback error learning equation is able to quickly train the neurofuzzy controller to provide tight setpoint tracking. Another advantage is that the neurofuzzy controller that employs the proposed online learning mechanism can be commissioned easily.