This paper presents a once-differentiable control strategy for a class of uncertain nonaffine nonlinear systems based on self-structuring neural networks (SSNNs) approximation, such that the system output tracks the desired trajectory. The optimal weight for each neuron in current SSNN is time-varying signals factually, and current stability analysis is only fit for a dwell time. Current SSNN control laws are not smooth and even not continuous, due to addition or pruning of neurons in the approximation procedure. In this paper, a new SSNN estimator and a new weight update law are proposed to ensure the optimal SSNN weights being constant values and the control law being once-differentiable. The effectiveness of the proposed control law is illustrated by the stability analysis in the whole tracking procedure and shown by the simulation results.
Paper
Full text
Adaptive Control for Nonaffine Nonlinear Systems Using Reliable Neural Network Approximation
Semantic Scholar · Computer Science · 2017
Abstract
This paper presents a once-differentiable control strategy for a class of uncertain nonaffine nonlinear systems based on self-structuring neural networks (SSNNs) approximation, such that the system output tracks the desired trajectory. The optimal weight for each neuron in current SSNN is time-varying signals factually, and current stability analysis is only fit for a dwell time. Current SSNN control laws are not smooth and even not continuous, due to addition or pruning of neurons in the approximation procedure. In this paper, a new SSNN estimator and a new weight update law are proposed to ensure the optimal SSNN weights being constant values and the control law being once-differentiable. The effectiveness of the proposed control law is illustrated by the stability analysis in the whole tracking procedure and shown by the simulation results.