Composite Learning for Trajectory Tracking Control of Robot Manipulators with Output Constraints

Abstract

In this paper, a trajectory tracking scheme for robot manipulators with unknown dynamics is investigated, with the consideration of output constraints as well as small bounded external disturbances. Firstly, a modified backstepping control scheme is employed to control the robot manipulators where in the first step of the design a tan-type barrier Lyapunov candidate is chosen in order to tackle the constraint problem. Secondly, the philosophy of dynamic surface control is incorporated to implement the calculation of prediction errors, which can also reduce “explosion of complexity” of the backstepping scheme. In addition, composite learning is introduced for a better estimation of unknown parameters, and for canceling out the uncertainties of the robot manipulators. Stability analysis shows that the proposed control scheme guarantees a small bounded tracking error with parameter convergence in the absence of the stringent persistent excitation condition. Finally, a simulation is conducted and the results demonstrate the superiority of the proposed controller in the aspects of tracking capability and parameter estimation.