Neural network-based intelligent perception guaranteed performance control for mechanical arm

Abstract

We concern with the intelligent perception control of a mechanical arm dynamic system with unknown initial state values, parameter perturbations, and external disturbances. Unlike existing prescribed performance control (PPC) methodologies which fail to preset convergence time via parameter setting, we propose a new type of PPC with predefined convergence time to impose prescribed behaviors on angle tracking errors. To accomplish such aim, we firstly define a predefined time stability criterion with an upper bound of convergence time that can be set in advance, and then we further convert the actual tracking error variable into a new variable with an initial value of zero by utilizing the error conversion function. Furthermore, a boundary amplitude intelligent extension algorithm is designed based on tracking error for performance constraint function (PCF), and meanwhile the radial basis function neural network (RBFNN) is adopted to approximate the mechanical arm system model. On this basis, a new PPC approach guaranteeing predefined convergence time is addressed for the mechanical arm system. Finally, the obtained simulation results reveal that the angle tracking error always evolves inside the extended boundary of the PCF, to satisfy better prescribed transient and steady-state properties in comparison with existing technics.