ABSTRACT
Control and stabilization of mechanical systems with nonholonomic constraints has been an area of active research. Due to Brockett’s theorem, it is well known that nonholonomic systems with restricted mobility cannot be stabilized to a desired configuration via differentiable, or even continuous, pure-state feedback, although it is controllable. For the controller design of nonholonomic systems, there are efforts focused on the kinematic control problem, where the systems are represented by their kinematic models and the velocity acts as the control input. Due to its great learning capability, it can be used to approximate any continuous function to any desired accuracy. Different control performance can be achieved by adjusting parameter adaptation gains and other factors, such as the size of the networks, and the exploration of the knowledge of the systems. Numerical simulation has been carried out to show the effectiveness of the proposed method for uncertain mobile robots.
