A proposed controller strategy based on a fuzzyneural network controller coupled with a recurrentneural network identifier for nonlinear electronicthrottle control

Recently, the application of the electronic throttlehas been very popular in the automotive industry. However,difficulties in the control of electronic throttle valves exist due tomultiple nonlinearities and plant parameter variations. A neuralnetwork based self-learning control (SLC) strategy that consistsof a fuzzy neural network (FNN) controller and a recurrentneural network (RNN) identifier is proposed for electronicthrottle valves in this paper. The FNN controller, which combinesthe semantic transparency of rule-based fuzzy systems with thelearning capability of a neural network, is utilized as an SLCscheme and will be robust to plant parameter variations. AnRNN identifier is employed to model the plant and provides plantinformation for the learning of the FNN controller. Both thestructure and the learning algorithm of the control system arepresented. The proposed controller is verified by computersimulations and experiments. The controller and the identifiercan be updated online by using the backpropagation algorithm,which is aimed to enhance the robustness with respect to plantparameter variations. The presented SLC strategy hasexperimentally been verified. A backpropagation algorithm usingthe gradient descent method is employed to train the proposedcontrol strategy to minimize the error between the actual outputand the desired output. Recent results show that the FNN is apromising approach to reap the benefits of both fuzzy systemsand neural networks to solve their respective problems, and theFNN can adjust the parameter of the fuzzy rules using a neuralnetwork-based learning algorithm.

Keywords:

Electronic throttle valve , fuzzy neural network(FNN) , nonlinear control , recurrent neural network (RNN) , selflearning