Remarks on self-tuning feedback controller using the Clifford multi-layer neural network

Abstract

In this study, Clifford multi-layer neural networks using a back-propagation algorithm are applied to control a nonlinear dynamic system to investigate its capability in practical control applications. A self-tuning feedback controller in which feedback gain parameters are adjusted by the Clifford multi-layer neural network is designed and a trail-based learning architecture is introduced in the online drawback learning of the Clifford multi-layer neural network. Computational experiments using a cart and a pendulum system as a plant that is controlled by the self-tuning feedback controller are conducted. In particular, the Clifford multi-layer neural networks followed by the Clifford algebras C0,0, C0,1 and C0,2 are utilised in the self-tuning feedback controllers, and these control performances are compared. Experimental results show that the Clifford algebra framework is feasible for improving the efficiency of neural computing. Results also confirm the potential of the Clifford multi-layer neural networks in control systems.