Recurrent Neural Network Inspired Finite-Time Control DesignOACSTPCDEI

Dear Editor,This letter is concerned with the role of recurrent neural networks(RNNs)on the controller design for a class of nonlinear systems.Inspired by the architectures of RNNs,the system states are stacked according to the dynamic along with time while the controller is represented as the neural network output.To build the bridge between RNNs and finite-time controller,a novel activation function is imposed on RNNs to drive the convergence of states at finite-time and propel the overall control process smoother.Rigorous stability proof is briefly provided for the convergence of the proposed finite-time controller.At last,a numerical simulation example is presented to illustrate the efficiency of the proposed strategy.Neural networks can be classified as static(feedforward)and dynamic(recurrent)nets[1].The former nets do not perform well in dealing with training data and using any information of the local data structure[2].In contrast to the feedforward neural networks,RNNs are constituted by high dimensional hidden states with dynamics.