Adaptive neural network control of bilateral teleoperation with constant time delay

Abstract

This paper proposes a novel approach for bilateral teleoperation systems with a multi degreesof-freedom (DOF) nonlinear robotic system on the master and slave side with constant time delay in a communication channel. We extend the passivity based architecture to improve position and force tracking and consequently transparency in the face of offset in initial conditions, environmental contacts and unknown parameters such as friction coefficients. The proposed controller employs a stable neural network on each side to approximate unknown nonlinear functions in the robot dynamics, thereby overcoming some limitations of conventional controllers such as PD or adaptive controllers and guaranteeing good tracking performance. Moreover, we show that this new neural network controller preserves the control passivity of the system. Simulation results show that NN con-A. Forouzantabar ( ) troller tracking performance is superior to that of conventional controllers.

Introduction to Haptics for Neurosurgeons

“…57,58 A similar approach using neural networks to estimate both unknown local interface and remote robot dynamics can preserve the passivity of the system. 59 For adapting to unknown human and/or environment dynamics (in addition to unknown robot dynamics), one can assume that these dynamics are linear, and then separate adaptive laws can be designed for the local side and remote side. 60,61 Even the uncertain lengths of the tools can be included for adaptation to kinematics.…”

Improved adaptive neural network control for humanoid robot hand in workspace

“…Zhang et al 19 proposed a model-free fuzzy basis function network to approximate the nonlinear function of observer and controller, and compensate the effect of system deviations to the velocity reconstruction. Forouzantabar et al 20 proposed a controller that employed a stable NN on the master or slide side to approximate unknown nonlinear functions in the robot dynamics, which had advantages of overcoming the limitations of conventional controllers such as PD or adaptive controllers and guaranteeing good tracking performance. Yao et al 21 developed an NN adaptive inverse controller to an electro-hydraulic servo system based on the adaptive inverse control theory and NN, and the control scheme was verified capably of tracking desired signals with high accuracy and good real-time performance.…”

“…In the equation, the first three terms of the controller are the model-based control and the fourth term Kr with K 2 R nÂn 4 0 is equivalent to the proportion derivative (PD) control type, which is an effective introduction to the controller through the definition of r proposed in equation (20), and the last term…”

Data driven soft sensor development for complex chemical processes using extreme learning machine

“…However, the process operations in many industries especially Cybenko, 1989;Narendra and Parthasarathy, 1990); secondly, ANN has the ability to learn from the known information and then predict the unknown with certain accuracy; lastly, there is no requirements for the knowledge of the related processes when developing the black-box models using the ANN (Almeida, 2002;Minns and Hall, 1996;Mjalli et al, 2007). What is more, the intelligent control capabilities for complex processes can be remarkably improved with the aid of soft sensors based on ANN models (Forouzantabar et al, 2012;Gonzaga et al, 2009;Huang and Lewis, 2003;Tong et al, 2011). Back propagation (BP) network is one kind of the most widely used neural networks (Chen, 1990;Heermann and Khazenie, 1992).…”