Nonlinear multivariable adaptive control using multiple models and neural networks

Data-Based Virtual Unmodeled Dynamics Driven Multivariable Nonlinear Adaptive Switching Control

“…Using the similar formulation to those in [7] the above results can be proved. This is omitted here.…”

“…The simulation results demonstrate the superior performance of the proposed approach over the results in [5]- [9]. However, due to the space limitation, in the following, we only show the comparison results with [7], which is the benchmark in this research area. In simulations, the two layer BP neural networks have been used to approach v(k).…”

“…For a class of nonlinear multivariable systems with unknown parameters, an adaptive decoupling control that combines a generalized predictive control with neural network compensation has been developed in [5], where global convergence analysis was not discussed. Using the method in [6], a stable and convergent multivariable adaptive control method was established in [7] for a class of unknown multivariable nonlinear system with stable zero dynamics, where the system was expressed by a combination of unknown linear subsystem and high order nonlinear terms. For a class of unknown multivariable nonlinear systems with unstable zero dynamics, a neural network and multiple-model based dynamic decoupling control algorithm was proposed in [8], where it is shown that the resulting control algorithm is stable and convergent when the k-difference for the high order nonlinear term is globally bounded.…”

Function Approximation-based Sliding Mode Adaptive Control for Time-varying Uncertain Nonlinear Systems

“…2005; Fang et al, 2006;Chiang et al, 2007), neural network technique (Yang & Calise, 2007;Fu & Chai, 2007;Zhou et al, 2007Tang et al, 2007 and fuzzy method (Hsu & Lin, 2005;Huang & Chen, 2006;Liu & Wang, 2007). For instance, combining a linear nominal controller with an adaptive compensator, Ruan (2007) and Hovakimyan (2006) realized the high performance stabilizing of inverted pendulum with un-modeling nonlinear dynamics.…”

A Multimodel Approach for Complex Systems Modeling based on Classification Algorithms

“…Many authors propose to apply classification algorithms in order to handle a set of dynamical models. For example, neural networks have been used to represent and control complex systems [7,[12][13][14][15]. In another hand, thanks to their ability to classify data and their simplicity, K-means algorithms have proved to be efficient for data clustering e.g.…”