NEURAL NETWORK-BASED ADAPTIVE TRACKING CONTROL FOR A NONHOLONOMIC WHEELED MOBILE ROBOT SUBJECT TO UNKNOWN WHEEL SLIPS

摘要

In this paper, Lagrange formula is employed with the purpose of modelling the both kinematics and dynamics of a nonholonomic wheeled mobile robot (WMR) subject to unknown wheel slips, model uncertainties such as such as unstructured unmodelled dynamic components, and unknown external disturbances such as unknown external forces. Afterwards, an adaptive tracking controller based on the radial basis function neural network (RBFNN) with an online weight tuning algorithm is proposed for tracking a predefined trajectory. The online weight tuning algorithm is modified from the backpropagation plus an e-modification term required for ensuring that the weights are bounded. Preliminary neural network offline training is not needed for the weights since they are easily initialized. Thanks to this proposed controller, a desired tracking performance is obtained in which not only position tracking errors uniformly ultimately converge to an arbitrarily small neighborhood of the origin but also the RBFNN weights are bounded. In the sense of Lyapunov and LaSalle extension, the stability of the whole closed-loop system is guaranteed to achieve this desired tracking performance. The result of computer simulation has validated the rightness and efficiency of the proposed controller.