Variable Structure Rough Neural Network Control for a Class of Non-Linear Systems

In this paper, a novel rough neural network control system based on the variable structure control developed for a class of SISO canonical nonlinear systems with taking the presence of bounded disturbance into account is presented. We assume that the nonlinear functions of the system are completely unknown. The rough neural network presented here is used to approximate the unknown nonlinear functions to a desired appropriate approximation. A fuzzy soft switching structure is developed to decide the amount of efforts taken by neural network and variable structure control systems based upon the real-time error characteristics.  A proper Lyapunov function is defined and used to deduce adaptive laws for tunable parameters of neural network and to achieve the closed loop stability of overall system. The rough family of neural networks have a reputation of better functionality at the presence of noise and disturbance, which comes from their interval characteristic of their parameters. In this study, we utilize this property to achieve better performance. To demonstrate the effect of proposed control structure, it is applied upon three systems (one exemplary system, a dynamical and a chaotic) and the simulated results show the efficiency of this hybrid variable structure control scheme.