Neural network adaptive control of a deployable manipulator

This paper presents an effective neural network based adaptive controller for a newly designed manipulator that has a deployable link as well as a revolute joint. The prototype manipulator system is described. The analytical formulation of the system is presented for the purpose of effective control. The relevant techniques of adaptive control of robot manipulators are presented. A single-layer, linear-in-the-parameter neural network that is based on Gaussian radial basis functions is used to approximate the unknown terms in the dynamical equations of the manipulator. The Lyapunov stability analysis is used to find an adaptive update rule for tuning the weights of the neural network. The corresponding adaptive controller is derived based on this approach. The applicability of the control scheme for this manipulator system is tested through computer simulations.