A training rule which guarantees finite-region stability of neural network closed-loop control: an extension to non-Hermitian systems

A training rule for neural network controllers that guarantees finite-region stability of control systems has previously been developed. The training rule requires that the controlled systems must be locally Hermitian and controllable and full state accessible. The controller is a single hidden layer feedforward network. This paper extends the training role by modifying the stability condition to drop out the Hermitian requirement. A finite stability region is estimated by evaluating an existing Lyapunov function, which is a by-product of the training rule