Title: Fixed final time satellite attitude control with thrusters based on dynamic programming and neural networks
Authors: Amin Ghorbanpour
Addresses: Department of Mechanical Engineering, Cleveland State University, Cleveland, OH, USA
Abstract: This paper studies the attitude control of a satellite in three-axis using only four thrusters. The mathematical model of the attitude is represented as a switched system with 16 subsystems. Each subsystem is defined based on the thruster's on/off status. The dynamic programming is utilised to find the optimal switching between subsystems such that a cost function is optimised. Furthermore, to generalise the solution for a specific domain of interest, neural network is employed for offline training and approximating the cost function. An offline training algorithm is suggested to find the optimal weights of the neurons and determine optimal switching. It is shown that the proposed method can execute a manoeuvre within fixed final time. Moreover, the control is robust against uncertainties in the system modelling. Finally, the system modelling and control approach is suggested as a framework to design low-cost attitude control system unit, suitable for microsatellite-class.
Keywords: satellite; attitude control; thruster; dynamic programming; machine learning; neural network.
International Journal of Space Science and Engineering, 2021 Vol.6 No.3, pp.257 - 274
Received: 25 Jul 2020
Accepted: 27 Oct 2020
Published online: 16 Mar 2021 *