On the Optimality of Fazzy PID Neuro-Gain-Scheduling Controller for Nonlinear Svstems

A novel optimized fuzzy PID neuro-gain scheduling controller is proposed in this paper A ner.tral netw ork adaptively assigns the appropriate paratneters for the fuzzy PID controller. Genetic algorithms are used to find the optimum training data for the neural network. Also, GA is thereafter once again used to optimize the neural network parameters such as number of neurons, layers and activation functions so that to minimize a predefined enor Jimctions. The fuzzy controller is designed according to human expertise. As a case study, the proposed controller is applied on a nonlinear robot arm with a one degree freedom and compared with classical gain schectuling and a conventional fuzzy neuro-gain schedtling controllers. Simulation results show that optimizing the parameters of the neural network( such as number of neurons and layers) with genetic algorithms, can greatly enhance the performance ofthe controller while its implementation complexity remains the same or even decreases.