Neural network-based integral sliding mode control of arbitrary nonlinear systems with unknown bounded disturbances

In this paper, a new technique of online learning radial basis function neural network-based integral sliding mode control for unknown nonlinear systems is proposed. The proposed controller is investigated in the presence of unknown time-varying and different types of disturbances. In this work, The weights of the designed isolated radial basis function neural networks are updated based on the derived adaptation algorithm to approximate the nonlinearities of the system. Sliding mode control helps to eliminate remaining approximation errors by neural networks. The integral sliding surfaces are newly designed, which helps control the system in the presence of non-smooth disturbances. Well-known nonlinear systems, namely the Duffing oscillator and the Lorenz system with special characteristics are considered here for a demonstration of the theoretical development. Chaotic and periodic properties of the Lorenz system are considered and the proposed controller has successfully stabilised the system with arbitrary initial conditions in the presence of unknown smooth and non-smooth disturbances. The superiority of the proposed neural network-based integral sliding mode control is established by comparing it with three different control techniques, i.e. proportional integral sliding mode control, terminal sliding mode control, and neural network-based sliding mode control.