Self-learning Neural Network Control System for Physical Model with One Degree of Freedom of System of Active Vibration Isolation and Pointing of Payload Spacecraft

Abstract

Purpose The area of biomimetic robots is successfully developing in intelligent robotics using SEMS and Neurotechnology. These robots are based on the borrowing its core elements from nature and able to adapt to the environment of the real world and to be truly intelligent autonomous robotic devices. For example, the neural network control system are used in intelligent robots, capable of self-learning like brain. Overall, the self-learning neural network control system have a structure similar central and peripheral nervous systems of vertebrates and man. The aim of the publication is the description of the developed model of the self-learning neural network control of a single-stage physical model of intelligence system of active vibration protection and very precise pointing of large precision space antennas. Results Model of the self-learning neural network control system of a single-stage physical model of intelligence system of active vibration protection and very precise pointing of the payload of the spacecraft is developed and tested. The advantages of application of neural PID controller are shown compared with conventional PID controller. Practical value The presented in the article the self-learning neural network control system of a single-stage physical model can be used to create autonomous intelligent robotic system capable to react to changing uncertain conditions in real time outside the operator’s actions, for example in deep space.