Abstract
Hitherto, all the existing techniques for controlling nonlinear systems required at least one of the following assumptions: the system is affine, is SISO, has a known nonlinear structure parametrized linearly by unknown parameters, and full state feedback is possible. This paper proposes a technique for the control of general nonlinear systems with unknown structure, which requires none of the above assumptions. The dynamic control law is based on the implicit function theorem, where the underlying unknown implicit function is emulated by a feedforward neural network. The controller can either be trained offline or adaptively trained in a direct adaptive control mode. Simulation examples are presented to illustrate the effectiveness of the proposed technique.