The main contribution of this paper is to realize a stability-based model-free control approach based on a conservative online-judgment of measured states with several assumptions. This new type of adaptive control with embedded stability-based active control adaption is based on a cognition-based framework, which consists of three parts: (1) a dynamic recurrent neural network (DRNN) used for local identification and multi-step-ahead prediction of the system; (2) a geometrical criterion based on a suitable definition of quadratic stability applied for judging the stability of motion of the system numerically; (3) a suitable strategy according to a cost function used for choosing most suitable control input value for the next predefined time interval. The proposed controller is able to gain useful local knowledge and define autonomously suitable local control input according to the stability criterion. Numerical examples are shown to demonstrate the successful application and performance of the method.

This content is only available via PDF.
You do not currently have access to this content.