Design A RBF neural network auto-tuning controller for magnetic levitation system with Kalman filter

Abstract

In this paper, a new controller structure called the radius basis function (RBF) neural network auto-tuning PID controller with Kalman filter is presented to manipulate a linearized magnetic levitation system. The proposed RBF neural network auto-tuning PID controller with Kalman filter makes use of Kalman filter to deal with the uncertainties and noises induced by the process of linearization of magnetic levitation system as well as the noise problems induced by the position feedback sensor device. To validate the proposed new design structure, the MATLAB simulations under different types of noise problems are presented. Furthermore, results confirm that the output transient response and steady-state error of magnetic levitation system by the proposed controller with Kalman filter can be improved and assured while the results of auto-tuning PID controller are inclined to be unstable.

Biologically-Inspired Learning and Adaptation of Self-Evolving Control for Networked Mobile Robots

“…Combining Equations (12) and (13), the kinematics of the four-wheeled omnidirectional mobile robot is derived as follows:    .…”

“…The Kalman filter is a state estimation technique introduced by R.E. Kalman [13]. It is a classic state estimation technique used widely in engineering applications, including spacecraft navigation, motion planning in robotics, signal processing and wireless sensor networks [14][15][16] because of its ability to extract useful information from noisy data.…”

“…It is a classic state estimation technique used widely in engineering applications, including spacecraft navigation, motion planning in robotics, signal processing and wireless sensor networks [14][15][16] because of its ability to extract useful information from noisy data. It is an optimal estimator for evaluating the internal state of a dynamic system under certain process patterns and/or measurement disturbances in the physical environment [13,17,18]. The objective of the Kalman filter is to minimize the mean squared error between the actual and estimated data.…”

“…Figure 1 presents the structure of Kalman filter algorithm in which the estimated state ˆ( 1) x k − and the error covariance P(k − 1) are included [13]. As shown in Figure 1, The Kalman filter algorithm consists of two phases: time update phase (predictor) and measurement update phase (corrector).…”

The feasibility of designing a back propagation neural network to predict the levitation force of high-temperature superconducting magnetic levitation

“…In [27], a deep learning controller based on a deep belief network was designed for the EMS Maglev train, and achieved better results than conventional PID control. In addition, an RBF neural network autotuning PID controller with Kalman filter was presented [28], and validated by MATLAB simulation. However, the research of HTS Maglev (especially the force prediction) based on the artificial intelligence (AI) technology has not been reported, and this intersection of superconductor and AI technology needs to be further investigated.…”