Adaptive neuro-sliding mode control of PUMA 560 robot manipulator

Abstract

The classical sliding mode control (SMC) is a robust control scheme widely used for dealing with nonlinear systems uncertainties and disturbances. However, the conventional SMC major drawback in real applications is the chattering phenomenon problem, which involves extremely high control activity due to the switched control input. To overcome this handicap, a pratical design method that combines an adaptive neural network and sliding mode control principles is proposed in this paper. The controller design is divided into two phases. First, the chattering phenomenon is removed by replacing the sign function included in the switched control by a continuous smooth function; basing on Lyapounov stability theorem. Then, an adaptive linear neural network, that has the role of online estimate the equivalent control in the neighborhood of the sliding manifold, is developed when the controlled plant is poorly modeled. Simulation results show clearly the satisfactory chattering free tracking performance of proposed controller when it is applied for the joints angular positions control of a 6-DOF PUMA 560 robot arm.