Formation control, manipulator control and soft robot based on neural network control system

Abstract

Listen

Translate article

Recent years, significant efforts for the applications of neural network on robotic adaptive control system are presented in many good quality research papers. Because neural network can learn to adapt to environmental changes and approximate functions, it is widely used in control systems. In order to further develop intelligent control, the application of learning neural network in control system is very important. In this paper, variety of schemes and molds on applications of N eural Networks (NN s) used in robotic area are reviewed and discussed in order to have an outlook of how Neural Network have been utilized in control of robots. By analyzing the existing research, the author found that the neural network has been well developed in the control system and used as a function approximator as usual. In fact, neural networks are sometimes used to control a single part of the system. For example, visual recognition using convolutional neural network in formation control, neural network as a general approximator to deal with the uncertain nonlinearity or uncertainty of the control system, such as radial basis function neural network (RBFNN) and so on. ThenDa new soft climbing robot based on Neural Networks, which is designed by author, is discussed in this paper to show some future directions of robotic control system. The robot has the ability of self-learning and adapting to the environment. It can judge its state according to its environment, and use neural network to learn and remember the new steady state for subsequent state judgment. Above all, several types of NNs with different features and a new designed soft robot are introduced in this paper in terms of their playing roles in robot formation control and robot manipulator control system, and some of the background knowledge related to certain kind of NN s are also introduced briefly.